Acknowledgements There are many people from whose help and comments I have benefitted considerably. First and foremost, I would like to thank Jürgen Esser for his advice throughout the project and his support way beyond the call of duty, without which this book would never have materialised. To him this book is dedicated. I would also like to thank Manfred Kohrt, Winfried Lenders, Karl Reichl and Klaus Peter Schneider who read earlier drafts of the manuscript and provided me with detailed feedback. For comments and discussions on individual examples and analyses, arguments and conclusions, I am grateful to Bas Aarts, Jan Aarts, the late Ruth Brend, Sylviane Granger, Sebastian Hoffmann, Rolf Kreyer, Geoffrey Leech, Jacqueline Monschau and the audiences at various conferences, in particular at the LACUS Forum 2001 in Montreal and the ICAME Conference 2002 in Gothenburg. Special thanks are due to my native-speaker informants Anne Barron, Rosemary Bock, Shalini Gupta, Sach Mukherjee and Sally Schmiesing. I would also like to express my thanks to the series editors for including this book in Language and Computers. Christian Mair in particular gave me invaluable advice at various stages of the editorial process. The proof-reading of the final version of the manuscript was taken over by Rosemary Bock, for which I am most grateful. Naturally, I remain responsible for all remaining blunders and infelicities. Finally, I owe an extreme debt of gratitude to my wife, Nicole, for her unfailing support and love.
Chapter 1 Ditransitive verbs in previous research and in the present work In spite of the vast literature that ditransitive verbs have spawned, linguists still do not unanimously agree on what ditransitive verbs are. Even the viability and scope of the concept of ditransitivity in general are a matter of dispute. Presumably this is why there are so many different models for the description of ditransitivity. Thus, an overview of various existing models is an appropriate starting-point for the present study. 1.1
Aims of the present study
The differences between existing models of ditransitivity can be systematised according to various dimensions. For instance, some linguistic approaches to ditransitivity are inherently lexical and focus on lexical features of ditransitive verbs (e.g. in terms of valencies as in valency theory, see section 1.2.3), while other models place special emphasis on the syntactic nature of ditransitivity (e.g. in terms of the ditransitive construction as in construction grammar, see section 1.2.7). Furthermore, models of ditransitivity may be functionalist (as in systemicfunctional grammar, see section 1.2.4) or formalist (as in generative grammar, see section 1.2.2). Also, different models have been derived from different kinds of data, drawing on different methodologies. With regard to the quantity of data, the corpus revolution has no doubt led to “data resources never available to linguists before” (J. Aarts 2000: 35), which provide an unprecedented mass and variety of empirical evidence for or against the plausibility of descriptive models. But even if we confine ourselves to linguistic research before the advent of modern, computerised corpora, it is obvious that fundamentally different kinds of data have been used. On the one hand, there is a well-established tradition of introspection and the use of invented and decontextualised sample sentences (most prominently in formalist schools such as generative grammar). On the other hand, functional grammarians have always placed special emphasis on the context of language use and, thus, on the use of attested and authentic language data (e.g. in British contextualism). Mention should also be made of completely different objects of inquiry in different linguistic schools: while, for example, generative grammar and cognitive grammar have been concerned with abstract competence, functional grammar and corpus linguistics have been interested in language as performance. Because of such different perspectives, methodologies and databases in the linguistic sciences, it does not come as a surprise that a multitude of models compete with each other for the best account of ditransitivity. Before I present my
2
Chapter 1
own approach, it is therefore necessary to critically review existing linguistic frameworks in order to identify useful points of reference on the one hand and crucial shortcomings on the other. Such a comparative analysis of major theories of ditransitivity is still lacking. I feel that linguistic research is often too closely associated with a particular linguistic school of thought from the outset, whereas it would be useful to integrate concepts and suggestions made in other linguistic schools. Therefore, the present chapter provides an overview of several relevant theories and identifies important conceptual overlaps and descriptive correspondences. In this, I regard the theoretical and methodological diversity in previous research into ditransitive verbs as a goldmine for a pluralist theory that should unite relevant aspects of different theories and combine them with an empirically sound methodology. As pointed out by Halliday et al. (1964: 301) and Stubbs (1986: 3), theoretical pluralism is needed because different linguistic models may account for different facts in language, depending on the specific descriptive goal and purpose. Thus, a pluralist framework, which is intended to account for various interrelated facts of a linguistic phenomenon, ought to integrate plausible concepts from various frameworks without being biased towards a specific model from the outset. More specifically, the pluralist theory of ditransitivity that will be described at the end of this chapter (and that will be at the basis of the present study) will be lexicogrammatical, semantically-oriented and functional. The overarching plan of the present study is as follows: • Firstly, the present chapter is intended to assess the viability of existing accounts of ditransitive verbs (see section 1.2) and to derive from this analysis a pluralist theory that makes it possible to come to grips with the phenomenon of ditransitivity more comprehensively than before (see section 1.3). • Secondly, this theory will be combined with modern corpus-linguistic methodology. The methodology – and some of its theoretical implications – will be discussed in Chapter 2. Chapter 3 will then provide a corpus-based description of ditransitive verbs in language use. • Thirdly, I will suggest a framework in which the traditionally established gap between competence-related models of language cognition and performancerelated descriptions of language use can be bridged. In particular, I will argue that corpus evidence not only tells us important things about actual language use, but also about the cognitive entrenchment (i.e. speakers’ linguistic knowledge) of ditransitive verbs. In chapter 4, some major aspects of such a usage-based model of speakers’ linguistic knowledge of ditransitive verbs will be discussed and exemplified. • Finally, chapter 5 will provide a summary of the results and conclusions of the present study. It will be rounded off by suggestions for important future research.
Ditransitive Verbs in Previous Research and in the Present Work 1.2
3
Previous approaches to ditransitive verbs: an overview and a critical review
I will start off by discussing some important descriptive grammars of English, representing the – largely European – structuralist tradition in its widest sense (see section 1.2.1). Afterwards, the explanatory power of the American generative paradigm will be explored (see section 1.2.2). Sections 1.2.3, 1.2.4, 1.2.5 and 1.2.6 are devoted to different perspectives on grammatical structures and relations which are more restricted in scope, that is, valency theory, functional grammar, corpus-based grammar and corpus-driven lexicogrammar respectively. Finally, I will discuss findings from more cognitive approaches, namely construction grammar and cognitive grammar (see section 1.2.7) and experimental cognitive sciences (see section 1.2.8). At the end of this chapter, all linguistic frameworks and models will be evaluated with regard to the main theoretical and descriptive objectives of the present work. It will emerge that, generally speaking, a modified version of the structuralist framework, combined with a corpus-based and empirical methodology, will be pursued in the present study. As will become clear from the following sections, this non-generative descriptive apparatus can be easily reconciled with plausible assumptions and results in cognitive linguistics. 1.2.1 Descriptive grammar The classic reference grammars of the early 20th century, in particular Poutsma (1904-1926), Kruisinga (1909-1932) and Jespersen (1909-1949), mark the beginning of a structurally-oriented, descriptive English grammar, replacing the then prevailing prescriptive tradition. Numerous concepts, terminologies and analyses of grammatical phenomena and constructions in the more recent grammars of the ‘Quirk fleet’ (Görlach 2000: 260) can be traced back to those classic reference grammars. It is therefore entirely appropriate to begin with an examination of the ways in which early descriptive grammarians tried to come to grips with ditransitive verbs. Jespersen (1927) does not explicitly use the term ‘ditransitive’, but describes the corresponding verb class as follows: Some verbs frequently or even regularly have two objects; we shall first mention the type: he gave the boy a shilling. Here it is customary to speak of the boy as the indirect, and a shilling as the direct object. (Jespersen, 1927: 278) Here, it becomes obvious that ditransitive verbs do not necessarily require two objects. As Jespersen puts it, they are only “frequently or even regularly”
4
Chapter 1
complemented with two objects.1 Thus, Jespersen implicitly raises the question of frequencies in authentic language use (say, with regard to the ditransitive use of the verb give), but as a matter of fact, an empirically sound answer to that question can only be obtained from large, machine-readable corpora by drawing on modern corpus-linguistic methods. Furthermore, what Jespersen (1927: 278ff.) has to say about verb complementation is interesting for two reasons: (1) his discussion of the suitability of the term ‘dative’ for present-day English; (2) his description of the tophrase which may replace the indirect object. These points deserve more detailed treatment. Jespersen (1927: 278) rejects outright the continued use of the term ‘dative case’ for a grammar of modern English, largely due to language-historical reasons: “it is just as unhistorical as it would be to speak of Normandy and New England as parts of the British Empire.”2 Nevertheless, he insists on the point that the distinction of direct and indirect object remains relevant to modern English grammar: Still, there is a real distinction between the direct and the indirect object in present day English. One important sign of the distinction is the possibility of substituting a to-phrase for an indirect object ... . (Jespersen, 1927: 279) In addition, Jespersen mentions further criteria for a clear differentiation between direct and indirect object. Of particular importance here is the fact that a ditransitive verb, e.g. offer, can be used with the direct object alone (e.g. they offered a reward), but not with the indirect object only (e.g. *they offered the man). The to-phrase and its systematic correspondence to the indirect object is also a matter of detailed discussion in Jespersen’s grammar. He argues that the indirect object is defined by its position before the direct object (and immediately behind the verb). Jespersen (1927) thus draws the conclusion that if the indirect object – by virtue of its realisation as a to-phrase – follows the direct object, one should no longer speak of an indirect object: In spite of the fact that “I gave the boy an apple” and “I gave an apple to the boy” are practically equivalent, it would be wrong to say, as is often said, that the boy and to the boy are the same “case” (dative) or 1
As far as the issue of frequency is concerned, Kruisinga (1925) is much less exact. However, he observes that, unlike the direct object, the entity in indirect-object position habitually belongs to an animate gender class: “Some verbs are used with two objects. In that case the first (indirect) object nearly always denotes a person ..., rarely a thing ... . The second (direct) object is usually non-personal, but it may denote a person ... .” (Kruisinga, 1925: 158). 2 Note that in generative grammar, the concept of ‘dative case’ is still implicitly present (most notably in notions such as ‘dative movement’, see section 1.2.2).
Ditransitive Verbs in Previous Research and in the Present Work
5
that to the boy is a dative-equivalent. Two constructions may mean the same, or nearly the same thing, and yet be grammatically different ... . The to-phrase is placed in another relation to the verb than the indirect object ... . (Jespersen, 1927: 291f.) Note that the attempt is made here to establish a clear boundary between syntax and semantics (which, as will be discussed later in this section, appears to be doomed to failure from today’s functionalist and corpus-linguistic perspective). Jespersen is certainly right in calling into question the appropriateness of the term ‘dative’ for present-day English. However, the to-phrase can still be seen as an ‘object’, which represents a predominantly semantically-defined functional category. I would contend that any definition of ditransitive verbs that is based on functional categories should not restrict the term ‘object’ to a specific formal realisation of objects (without a preposition, that is). It seems to be doubtful whether one should really insist on a rigid separation of syntactic relations from their implicit semantic values, as envisaged by Jespersen. The merely syntactic definition of the indirect object leads Jespersen (1927) to establish a separate type of ditransitive complementation with two direct objects: ask John ask a question
ask John a few questions
... it is therefore better here, where the to-phrase cannot be used, and where each of the two objects can stand by itself without the other, to speak of two direct objects. (Jespersen, 1927: 295f.) To regard both objects in such constructions as direct objects is a somewhat peculiar solution (and has, to my knowledge, not attracted wide-spread attention). If the functional categories of direct and indirect object are taken as primarily semantically defined categories, there is no reason why John (as the ‘affected entity’) and a few questions (as the ‘transferred entity’) could not be labelled as indirect object and direct object respectively. Furthermore, the question arises whether and to what extent the two objects of ask may really stand on their own, which is the main argument underpinning Jespersen’s hypothesis of two direct objects. It seems much more plausible to me to analyse the phenomenon at hand differently. I would argue that, in principle, the ditransitive verb ask requires two objects (which correspond to two different entities in the outside world, i.e. the question and the person being questioned). If, however, an object is inferrable from the context (situational or textual), it need not be mentioned explicitly. Consequently, the seemingly ‘monotransitive’ use of ask in ask John would then be interpreted as an ellipsis with the omission of the direct object. In Jespersen’s
6
Chapter 1
decontextualised examples, this hypothesis is not testable – but it remains a viable alternative interpretation.3 Unlike Jespersen, Kruisinga (1925) does go into details about the usual semantic roles of the indirect object: The indirect object usually expresses the person or thing that is benefited by the action (e.g. do, spare, allot), often combined with direction of place (e.g. bring, less clearly give) ... . (Kruisinga, 1925: 159) Additionally, two further aspects of Kruisinga’s description are particularly noteworthy because they can, by and large, still be found in the much more recent Quirk-grammars: (1) the concept of ‘prepositional object’; (2) the clausal realisation of the object (and the subject). That ‘plain objects’ (i.e. without initial preposition) and so-called ‘prepositional objects’ are in fact quite similar syntactically and semantically is described by Kruisinga (1925) as follows: Some prepositional adjuncts present peculiar features. When we say I laughed at him we have indeed a prepositional adjunct at him as far as form goes. But it is evident that the verb is not laugh but laugh at. The form him is, therefore, in much the same relation to the verb as in I saw him. This is also shown by the fact that him can become the subject in both cases: He was seen; He was laughed at ... . Hence these adjuncts are called prepositional objects. The prepositional objects are identical in function with what may be called the plain objects ... . The same distinctions with respect to meaning may be made here. (Kruisinga, 1925: 164) Kruisinga thus suggests that a prepositional phrase can, in principle, fulfil the same syntactic function as a noun phrase. With regard to the alternation between ditransitive variants such as I gave the boy the penny and I gave the penny to the boy, however, Kruisinga does not label the to-phrase as a prepositional object, but as a ‘an adjunct’ that is ‘equivalent’ to the indirect object (Kruisinga 1925: 160). This is only a small step away from calling the to-phrase in I gave the penny to the boy a prepositional object too – a step that Quirk et al. (1985) do take in the Comprehensive Grammar, as will be discussed later on. Kruisinga (1925) also explicitly states that the direct object, but not the indirect object, may be realised as a clause (Kruisinga 1925):4 3
This hypothesis is very much in line with Matthews’s (1981: 125f.) concept of ‘latent’ objects, which will be taken up in section 1.2.4. 4 Kruisinga’s explanation of why indirect objects cannot be realised as clauses remains unclear. As a matter of fact, one can easily think of clausal realisations of the indirect object as well, e.g. in She gave whoever entered the room a winning smile.
Ditransitive Verbs in Previous Research and in the Present Work
7
Object clauses: I expect he will come. I laughed at what he said. ... English has no clauses serving the function of an indirect object, because the relative pronoun, which would open such a sentence, is never used at the beginning of a sentence ... . (Kruisinga, 1925: 188f.) Even though Kruisinga (1925: 187) only allows for finite object clauses functioning as direct objects, his categorisation is innovative in that objects – defined semantically as well as syntactically – are now taken to be associated with a wide range of potential formal realisations, which is particularly relevant to the analysis of ditransitive complementation. The attempt to systematically integrate different formal realisations on functional grounds has exerted an enormous influence on structuralist grammars. This key concept is also at the basis of Quirk et al.’s (1985) Comprehensive Grammar – the culmination of the so-called ‘Great Tradition’ (F. Aarts 1975) –, to which I will now turn. With regard to ditransitive verbs, Quirk et al. (1985: 54) give the following, apparently clear-cut definition: “DITRANSITIVE VERBS occur in type SVOO”. This statement highlights the fact that ditransitive verbs are defined by means of the clause pattern in which they are used. It is the clause pattern in which both a direct and an indirect object (and a subject) are obligatorily required by the verb.5 Quirk et al. (1985: 56) give the following example: (1)
Type SVOO
S V (ditransitive) We all wish
Oi you
Od a happy birthday
The strictly clause-pattern-related definition of ditransitive verbs has clear advantages. First, whenever a verb requires a subject and two different objects, it is considered a ditransitive verb. Thus, a clear distinction is made between ditransitive verbs, bound to the clause pattern SVOO, and other trivalent verbs and clause patterns such as SVOC in which the complement is co-referential with the (only) object. Second, the definition of ditransitivity in terms of functional categories required by the verb makes it possible to cover a wide range of possible formal realisations of the functional category of (direct) object, e.g. noun phrase, wh-clause, that-clause. 6 Generally speaking, this consistent distinction of formal structures and functional categories seems to me to be one of the major 5
This is made even more explicit in the definition given in Quirk et al.’s (1972: 843) earlier grammar: “Ditransitive complementation involves two objects that are not in a coreferential, intensive relation”. This definition is clearly intended to demarcate the clause pattern SVOO from the clause pattern SVOC. 6 This general approach has not remained unchallenged. With regard to verb complementation, for example, Standop (2000) criticises many structural analogies drawn by Quirk et al. (1985). This criticism will be reviewed later in this section.
8
Chapter 1
strong points in the Comprehensive Grammar because it makes it possible to subsume different but analogous formal realisations into one and the same functional category. Before delving more deeply into Quirk et al.’s (1985) account of ditransitive verb complementation, it is necessary to point out two problems that are caused by their exclusively clause-pattern-related definition of ditransitive verbs.7 The following example sheds light on the first problem in the line of argumentation of the Comprehensive Grammar: (2)
May I inform you that your order is ready for collection?
Many verbs are said to be ditransitive which may occur in the clause pattern SVOO, but certainly not in the classic ditransitive realisation of the two objects as two noun phrases. Thus, Quirk et al. (1985: 1212) also ascribe the label ‘ditransitive’ to verbs such as inform which require an indirect object and a that-clause functioning as a direct object, cf. (2). This has to do with the fact that the notion of object in the Comprehensive Grammar is defined extremely broadly and refers to a wide range of formal realisations (from noun phrases to clauses) and even to prepositional objects. The solely clause-pattern-related definition of ditransitive verbs together with the broad understanding of the category of object automatically leads Quirk et al. (1985: 1211) to consider many verbs ditransitive which, in fact, cannot be complemented with two noun phrases (e.g. convince, inform) or which do not occur in the clause pattern SVOiOd at all, e.g. address to and communicate to. But even Quirk et al. (1985: 1208) themselves argue that the clause pattern SVOiOd with both objects realised as noun phrases represents the basic pattern of ditransitive verb complementation: “Ditransitive complementation in its basic form involves two object noun phrases …”. In my view, any definition of ditransitive verbs ought to take into account this basic formal realisation of the underlying clause pattern. This is not to say that many (if not all) verbs which are used ditransitively may not also occur in other types of complementation, especially in the monotransitive clause pattern. This phenomenon is described by Quirk et al. (1985) as the ‘multiple class membership of verbs’: It must be borne in mind that a given verb can belong, in its various senses, to a number of different classes ..., and hence enter into a number of different clause types. (Quirk et al., 1985: 720)
7
Note that (owing to this shortcoming, I suspect) it is not always clear whether the lexical verb in particular or the verb phrase in general is considered ditransitive. Usually, the Comprehensive Grammar makes implicit use of the former definition, but at times the complex verb phrase in its entirety is given the label ‘ditransitive’, for example the verb phrase must send in the example I must send my parents an anniversary card (Quirk et al. 1985: 721).
Ditransitive Verbs in Previous Research and in the Present Work
9
Besides the overextension of the concept of ditransitive verbs, a second problem stems from the fact that the Comprehensive Grammar largely neglects the semantics of ditransitive verbs and their corresponding clause pattern SVOO. Quirk et al. (1985: 740ff.) do account for ‘semantic roles of clause elements’ (e.g. with regard to the ‘affected indirect object’, which is particularly relevant to the issue at hand), but the general relations between syntactic categories and their semantic roles are not precisely mapped onto the different clause patterns in particular. However, ditransitive verbs share considerable similarities in their underlying propositions. In other words, the ditransitive clause pattern SVOO has an abstract meaning itself. This is pointed out, among others, by Jackson (1990), who sets out to develop a semantically-oriented syntax by explicitly drawing on the descriptive apparatus of the Comprehensive Grammar:8 The sixth pattern of complementation contains two Objects in addition to the Subject and the Predicator (6. SPOO), e.g. (two instances) [93] He – gives – the gen (and) we – give – him – the publicity [N04:131] The propositions expressed by sentences with this pattern commonly refer to accomplishments of transferring goods from one person to another. (Jackson, 1990: 159) A functional analysis of syntactic structures should take into account such correspondences between syntax and semantics whenever possible.9 As far as the complementation of ditransitive verbs is concerned, there is a general distinction of four main types of verb complementation in the Comprehensive Grammar: ‘copular’ [A], ‘monotransitive’ [B], ‘complex transitive’ [C] und ‘ditransitive’ [D]. With regard to the ditransitive type, Quirk et al. (1985) provide the following list of subtypes [D1] to [D6]:10 DITRANSITIVE (Type SVOO) [D1] Noun phrases as Oi & Od [D2] With prepositional O [D3] Oi + that-clause 8
They offered her some food. Please say something to us. They told me that I was ill.
Jackson uses ‘P’ (for ‘predicator’) instead of ‘V’ (for ‘verb’). The sentence-tag in brackets following the example refers to its position in the Lancaster-Oslo/Bergen Corpus of British English (LOB). 9 Recent corpus-linguistic studies provide ample testimony of the fact that such correspondences between syntactic patterns and their semantic values (which may be more or less abstract in nature) can be found in many fields of language use, which calls into question the widely held, formalist view of an autonomous syntax (see section 1.2.5). 10 Note that in many approaches to verb complementation in English the letter ‘D’ is used as an iconic symbol for the ditransitive type, e.g. on the back-cover of the Dictionary of Contemporary English (Procter, 1978). However, subtypes [D1], [D5] and [D6] to be found there are not identical with the subtypes in the Comprehensive Grammar.
10
Chapter 1 [D4] [D5] [D6]
Oi + wh-clause Oi + wh-infinitive clause Oi + to-infinitive
He asked me what time it was. Mary showed us what to do. I advised Mark to see a doctor. (Quirk et al., 1985: 1171)
These subtypes of ditransitive complementation are then discussed in detail and exemplified in the Comprehensive Grammar (Quirk et al., 1985: 1208ff.).11 Their account is surely the most elaborated attempt to characterise ditransitive verbs and their complementation from a descriptive and structuralist point of view.12 The overview of ditransitive complementation patterns above indicates that Quirk et al. pick up on Kruisinga’s wide understanding of objects (including plain objects, prepositional objects and object clauses). The Comprehensive Grammar thus systematises the variability of ditransitive complementation by drawing structural analogies (and, to a lesser degree, by sketching semantic similarities) between different formal realisations. Let us first turn to ditransitive complementations without clausal realisations of clause elements. It is subtype [D1], with both objects realised as noun phrases, which is usually considered the basic form of ditransitive complementation, as for example in She gave the girl a doll. According to Quirk et al. (1985: 1208f.), both the indirect object and the direct object may also be realised as prepositional phrases. If the indirect object is realised as a prepositional phrase, the clause element is called a ‘prepositional object’ (i.e. subtype [D2a] as in She gave a doll to her; but also in She addressed her remarks to the children, although this sentence has no [D1] counterpart). Similarly, the direct object may also be realised as a prepositional phrase (i.e. subtype [D2b] as in He reminded him of the agreement, which has no [D1] counterpart either). Subtypes [D2a] and [D2b], which include prepositional objects, are thus taken to be ditransitive complementation patterns because of structural analogies drawn between the basic form [D1] and subtypes [D2a] and [D2b] at the level of functional categories:
11 Also, some other marginal cases are covered, e.g. the subtype [D3pr], in which the verb is followed by a prepositional object and a that-clause, as for example in I ask of you that you will keep this secret (Quirk et al., 1985: 1213). 12 Not only does the Comprehensive Grammar provide the most comprehensive account of ditransitive verbs, but it also differs from previous grammars in that, for the very first time, the description of grammatical phenomena is complemented with authentic data obtained from large and representative corpora. However, the corpus data from the Survey of English Usage (SEU), the Brown University Corpus and the Lancaster-Oslo/Bergen Corpus (LOB) are not taken into consideration systematically, let alone statistically, so that the Comprehensive Grammar certainly does not represent a corpus-based grammar (cf. Sinclair’s (1991: 100f.) and Stubbs’s (1993: 9) critical remarks on the vague relation between the grammar and corpus data). Truly corpus-based grammars such as Sinclair (1990) and Biber et al. (1999) will be dealt with in section 1.2.5.
Ditransitive Verbs in Previous Research and in the Present Work (3)
a) She (S) b) She (S) c) She (S) d) He (S)
gave (V) gave (V) addressed (V) reminded (V)
11
the girl (Oi:NP) a doll (Od:NP) to her (Oi:PP) a doll (Od:NP) her remarks (Od:NP) to the children (Oi:PP) of the agreement (Od:PP) him (Oi:NP)
The analysis of the examples given in (3) makes it clear that the Comprehensive Grammar views the functional categories of direct and indirect object as independent of their formal realisations (as noun phrase or prepositional phrase) and of their position.13 The structural analogies sketched out in (3) are semantically motivated. This is the actual reason for the establishment of subtypes [D2a] and [D2b] as ditransitive complementation patterns. This functional approach, however, is strongly criticised, among others, by Standop (2000: 223), who argues that none of the sentences (3b), (3c) and (3d) represents ditransitivity of the verb. More specifically, he regards the prepositional phrase to her in (3b) as an obligatory adverbial, but not as an object of any kind. He also rejects the structural analogy that is drawn between (3b) and (3c) because, unlike sentence (3b), She addressed her remarks to the children in (3c) has no equivalent sentence with two noun phrases functioning as objects (that is to say, *She addressed the children her remarks is not possible). Accordingly, he does not subsume sentence (3d) into the ditransitive category either (again, because *He reminded him the agreement is not permissible). In effect, Standop’s line of argumentation boils down to a very narrow and exclusively syntactic definition of ditransitivity: he suggests that the label ‘ditransitive’ be used only for verbs which are complemented by two objects realised as two noun phrases (which is, by the way, in line with Huddleston and Pullum’s (2002) recent, more generatively-oriented Cambridge Grammar of the English Language, see section 1.2.2). In my view, Standop’s criticism of the classification scheme in the Comprehensive Grammar is not entirely implausible, but in the end unjustified because the fundamental differences in linguistic description between him and Quirk et al. are largely based on conflicting premises. While Standop (cf. 2000: 230) seeks to separate syntax and semantics altogether (thus suggesting entirely syntactic criteria for a categorisation of verb complementation patterns), the Quirk-grammars are in principle aimed at a functional description of syntactic relations which are always taken to be semantically motivated (perhaps most overtly in Leech and Svartvik’s (1975) Communicative Grammar). If we follow the assumption that syntax and semantics are virtually inseparable in a ‘grammar in use’, Standop’s remarks appear largely unsatisfactory. With regard to the object of inquiry at hand, it does not make any sense to me to consider give in 13
Note that Quirk et al. (1985: 59) also mention an alternative analysis along the lines of the clause pattern SVOA: “There is a further correspondence by which SVOO clauses can be converted into SVOA clauses by the substitution of a prepositional phrase following the direct object for the indirect object preceding it: She sent Jim a card ~ She sent a card to Jim.” However, in the section on ditransitive verbs, they stick to the analysis in which the prepositional phrases “are described as prepositional objects, and are regarded as grammatically equivalent to indirect objects.”
12
Chapter 1
(3a) a ditransitive verb, but not in (3b) just because the indirect object is now placed behind the direct object and realised as a prepositional phrase. The formal differences between (3a) and (3b) do not change anything about the fact that in both sentences, the verb give requires two obligatory objects, which correspond to semantically defined entities in the outside world.14 However, I agree with Standop that an overextension of systematic correspondences, resulting in the inclusion of a very heterogeneous group of prepositional verbs such as compare with, convict of and refer to (cf. Quirk et al. 1985: 1211) in the ditransitive type, is rather extreme and not very useful. However, my stance on this issue is not based on syntactic considerations, but on semantic ones. In my view, these prepositional verbs have nothing in common with the shared abstract meaning of genuinely ditransitive verbs, namely an underlying proposition denoting an event type in which a provided entity is transferred to an affected entity.15 It is therefore reasonable to ascribe the label ‘ditransitive’ only to those verbs which are attested in the basic form of ditransitive complementation, i.e. subtype [D1] in the Comprehensive Grammar. Once a given verb occurs in the basic form, variations of this basic complementation pattern are also considered ditransitive. For the purposes of the present study, this means that a verb is considered ditransitive whenever it occurs in the basic ditransitive complementation pattern [D1]. It is only then that all occurrences of the given verb in other ditransitive subtypes, i.e. [D2] to [D6], are also taken to be instances of a ditransitive complementation. This approach excludes all the prepositional verbs which the Comprehensive Grammar subsumes into the ditransitive category. On the other hand, a clear disadvantage is that verbs which are very similar to ditransitive verbs in terms of their syntax and semantics but which cannot be complemented by Oi:NP and Od:NP are also excluded. This, for example, holds true for the verb address in She addressed her remarks to the children. It goes without saying that the structural analogy between She gave a doll to her and She addressed her remarks to the children drawn in the Comprehensive Grammar (and criticised by Standop, see above) is certainly plausible from a semanticosyntactical point of view. Unfortunately, though, the form *She addressed the 14
Note that Standop (2000: 223) himself repeatedly discusses semantic aspects in order to clarify the boundaries of syntactic categories. For example, he refers to the example She gave a doll which is possible but has a meaning that is quite different from She gave a doll to her. From this semantic comparison, he then draws the conclusion that to her in She gave a doll to her is an obligatory clause element. It seems then that even for Standop syntactic analysis without consideration of semantics is not possible. 15 As will be discussed in section 1.2.7, this event type is, of course, a good example of ‘gradience’ in grammar (cf. e.g. B. Aarts, 2004), i.e. a category with central members, peripheral members and fuzzy boundaries. The most prototypical member of this event type would be a GIVE-event. More peripheral members would be events of the ENVYtype, which Gropen et al. (1989: 241) relate to the central GIVE-type by subsuming envy under verbs of ‘future not having’. Note that this extension of the ditransitive event type is only compatible with the double-object construction, i.e. subtype [D1], but not with the prepositional variant, i.e. subtype [D2].
Ditransitive Verbs in Previous Research and in the Present Work
13
children her remarks does not exist. One could hypothesise a ‘syntactic gap’ here. That is to say that the language system as such would allow for this formal realisation – because of the aforementioned structural and semantic analogies – but, for no apparent reason, it runs counter to the actual norm in language use and is thus not attested.16 Speaking of syntactic gaps, the ditransitive verb provide is an even more interesting example. Quirk et al. (1985: 1210) include this verb in the basic subtype [D1], but this usage is labelled ‘
’. From a large-scale analysis of four 1-million-word corpora and the 100-million-word British National Corpus (BNC), I have drawn the conclusion that in present-day American English, the verb provide is only used sporadically in subtype [D1] (cf. Mukherjee 2001: 299ff.) so that the question arises whether this complementation of provide is already a generally accepted and relevant norm in American usage. One could easily hypothesise in the light of the clear quantitative prevalence of subtypes [D2a] and [D2b] that there is a diachronic development going on in American English which could result in the establishment of the basic ditransitive complementation of provide, i.e. subtype [D1], in the foreseeable future. In other words, we are perhaps observing the closing of a syntactic gap as defined before.17 In a wider setting, this very example makes it clear that a categorisation of verbs into different types of complementation and/or transitivity which is also based on formal realisations will never cover all verbs which might fit in the categories from a semantic point of view. This disadvantage of introducing a formal component in the definition of ditransitive verbs is, however, largely outweighed by the advantage of not having to include verbs which do not at all meet the underlying semantic criteria of ditransitivity. Let us now turn to ditransitive subtypes [D3] to [D6] established in the Comprehensive Grammar. What they have in common is the realisation of the direct object as a finite clause (that-clause, wh-clause) or a non-finite clause (whinfinitive clause, to-infinitive). In particular, the special subcategory [D3pr] should be mentioned in which the ditransitive verb is said to be complemented with a prepositional object and a that-clause: Quirk et al. (1985: 1213) illustrate this pattern with constructions such as He wrote to me that ... and He reported to me that ... . It is obvious from Standop’s chain of argumentation outlined above that he calls into question the suitability of this category for a plausible linguistic description of ditransitive complementation. Since to me is a prepositional phrase, he does not consider such cases to be examples of ditransitivity (cf. Standop 16
In general, a ‘syntactic gap’ would then be a structurally possible syntactic form which, for whatever reasons, is not used. To a certain extent, such syntactic gaps resemble the ‘lexical gaps’ which Kjellmer (2003) has recently suggested. An example of a lexical gap would be the structurally possible but non-existent form *ancience, which might have been derived from the adjective ancient and which might mean “antiquity” or “ancient times”. 17 The plausibility of the hypothesis that the [D1]-complementation of provide is probably coming in was pointed out to me by Christian Mair (personal communication). To be precise, the [D1]-complementation of provide has probably been marginal for centuries, is steadily gaining acceptance in recent American English and is now spreading from there.
14
Chapter 1
2000: 254). As far as the clausal realisation of direct objects is concerned, note in this context that, unlike Quirk et al. (1985), F. Aarts and J. Aarts (1982: 138ff.) do not allow for a that-clause or a to-infinitive as direct objects since these constituents cannot become the subject of the corresponding passive sentence. I have already stated above that the present work is informed by the belief that Quirk et al.’s (1985) systematic correspondences between different formal realisations are in general useful (notwithstanding some problems involved, especially with regard to prepositional verbs) since they represent structural analogies which are symptomatic, as it were, of underlying semantic similarities. As will become clear at the end of this chapter, the present study is clearly to be seen in the descriptive tradition of the Quirk-grammars, but it is also an attempt to achieve a more balanced trade-off between syntactic and semantic criteria of ditransitivity. More specifically, I subscribe to the point of view that a definition of ditransitivity should unite syntactic and semantic aspects because the ditransitive syntax is indicative of ditransitive semantics. Thus, verbs should be considered ditransitive if they require a specific number and range of formal elements in order to attain syntactic completeness; the number and range of those formal elements, however, is largely conditioned by the meaning of the ditransitive verb. Thus, it makes sense to regard both I told him the story and I told him what happened as sentences with a ditransitive verb tell (which would be in line with the Comprehensive Grammar). The direct object may be realised as a noun phrase in the first sentence and as a wh-clause in the second one, but they are equally required by the ditransitive verb tell and have the same semantic role.18 This semantic analogy between the different syntactic structures is explained, for example, by Verspoor and Sauter (2000): Subjects and objects are the main participants in an event or situation and are usually persons or things. Therefore, they are commonly realized by a noun phrase. ... Sometimes a whole event or situation is seen as a 'thing' and can be seen as a participant. An event or situation is expressed by means of a clause. The type of clause that may function like an NP, often called a noun clause, may be finite or nonfinite. (Verspoor and Sauter, 2000: 154) In sum, there are many good reasons to base the description of ditransitive verbs on the categories and principles of the Comprehensive Grammar. However, in the present study the classification scheme in the Quirk-grammar is modified in one major respect since only those verbs are considered ditransitive which occur in the basic type of ditransitive complementation [D1] – with the formal realisation Oi:NP and Od:NP – in actual language use. If this is the case, all other variant complementations of a given verb are also taken to be ditransitive complementations of that verb. This procedure ensures that only verbs are taken into account 18
This is also very much in line with cognitive-linguistic concepts (see section 1.2.7).
Ditransitive Verbs in Previous Research and in the Present Work
15
which imply the ‘ditransitive meaning’, or – in Verspoor and Sauter’s (2000) terminology – display the ‘giving/buying pattern (ditransitive verbs)’: Sentences with the giving/buying pattern consist of a subject, predicator, indirect or benefactive object, and direct object. Therefore, for this pattern to occur, there must be an event involving at least three participants, a person who gives something to someone or does something for someone (the subject), then the thing that is given or done (the direct object), and the receiver (the indirect or benefactive object). (Verspoor and Sauter, 2000: 26) As will be shown later in this chapter, the notion of a ‘giving/buying pattern’ largely corresponds to descriptions of ditransitivity in valency theory and functional grammar (see sections 1.2.3 and 1.2.4). It also shows that the event type encoded by ditransitive verbs implies a transfer in the outside world which may be concrete or more abstract in nature. While a concrete transfer is linked to a proper ‘indirect object’, an abstract transfer is often associated with a ‘benefactive object’: The difference between an indirect object and a benefactive object is that an indirect object has the thing given in hand after the transfer, whereas the benefactive object does not … . John gave Peter the ball. The sentence implies that Peter now has the ball in his possession. John bought Peter the ball. This sentence does not necessarily imply that Peter now has the ball in his possession. (Verspoor and Sauter, 2000: 26f.) It will thus be necessary to include various types of transfer events in a model of ditransitivity so that both indirect objects and benefactive objects are accounted for.19 1.2.2 Generative grammar While descriptive and structurally-oriented grammars have paid particular attention to the systematisation of actual language use, the generative-transformational school of thought has exclusively focused on ideal speaker-hearers’ abstract competence, that is, on the development of a model of linguistic competence (or internal language, I-language). It is beyond the scope of the present work to review the history of generative approaches to language in its entirety. Rather, I am interested in the question whether the Chomskyan paradigm may offer more 19
Note again that the notion of transfer also includes instances of ‘negative transfer’, accounting for the ‘deprived’ object of verbs like envy.
16
Chapter 1
plausible and fine-tuned alternatives to the description and analysis of ditransitive verbs (and their complementation) than the descriptive grammars which have been dealt with in section 1.2.1. In this context, I will discuss in particular the most recent trend in generative grammar, for which the term ‘Minimalist Program’ (MP) has been coined (cf. e.g. Chomsky 1995, Radford 1997a and Radford 1997b).20 It should be noted, though, that despite the abundance of new concepts and terminologies, the Minimalist Program has to be seen in the tradition of previous generative theories such as the ‘Government-and-Binding’ (GB) and the ‘Principles-and-Parameters’ (P&P) theory in that all theories are intended to offer models of speakers’ abstract grammatical competence. In the Minimalist Program, the emphasis is shifted more towards ‘economy conditions’ so that the model can be represented as economically and with as few redundancies as possible. Thus, Culicover (1997) writes: The approach of MP is to account for the structure of language as the consequence of what are assumed to be intuitively natural ‘economy conditions’ on the computational mechanism that comprise grammars. (Culicover, 1997: 347) In pursuing the Minimalist Program, generativists dispense with a considerable number of concepts which were inherent in previous generative models. However, the subcomponents linked to those concepts are developed further and continue to be of pivotal importance for the minimalist approach, even though new technical terms are now introduced. For instance, the dichotomy of ‘deep structure’ and ‘surface structure’ is now extended by the ‘spell-out’ component, which makes it possible to proceed to the ‘phonetic form’ (PF) at any stage in the generation of language (Chomsky, 1995: 189). Also, the terminology of ‘government and binding’ is no longer used explicitly, but the underlying concept remains implicitly relevant (for example, when it comes to case assignment).21 In the Minimalist Program, a major “key assumption [is] that syntax is binary-branching” (Radford, 1997a: 379). Thus, any constituent is said to consist of a central ‘head’ and an extension of the head, i.e. the ‘specifier’. The constituent itself is the ‘projection’ of the head. Furthermore, a distinction is made only between two levels of projection. For example, in between the verb (V) and its maximal projection (the verb phrase, VP), there is only one additional level of projection, which is V'. The symbols for these categories (e.g. VP, V', V) are to be found at the ramifications and ends of the complex tree diagrams that evolve. Notwithstanding the similarities between minimalist tree diagrams and the diagrams in previous generative theories, there are two new categories in MP that are 20 Note, for example, that I am not going into details about Relational Grammar, the key interest of which is typological in nature: “in what ways do natural languages differ, and in what ways are they alike?” (Perlmutter, 1980: 195). 21 Interestingly enough, Chomsky (1995: 162), in reviewing previous generative theories such as the Government-and-Binding theory, speaks of a “misleading term that should be abandoned, in my view”.
Ditransitive Verbs in Previous Research and in the Present Work
17
particularly noteworthy: (1) the ‘determiner phrase’ (DP), replacing, by and large, the traditional concept of noun phrase (NP); (2) the ‘inflectional phrase’ (IP), which, among other things, is intended to replace the traditional notion of sentence (S) on top of a tree diagram. In fact, IP is a highly heterogeneous notion: “a category devised by Chomsky whose members include finite auxiliaries (which are INFLected for tense/agreement), and the INFinitivaL to)” (Radford, 1997b: 263). Figure 1-1, taken from Radford et al. (1999), visualises the way in which a simple sentence with a monotransitive verb (i.e. in the clause pattern SVO) is analysed minimalistically.
Figure 1-1:
An MP analysis of We do envy you (Radford et al., 1999: 316)
In Figure 1-1, IP comprises a determiner (D), which is the pronoun we functioning as subject, and I' (or Ī). I' comprises the auxiliary do, marked for tense and inflection (I), and a verb phrase (VP). Finally, VP consists of the lexical verb envy (V) and a determiner, which is the pronoun you functioning as direct object. Thus, the traditionally established unit of verb phrase is being analysed on two different levels: first, the lexical verb together with its object is considered a projection of the auxiliary do; second, the object itself is a projection of the lexical verb. Note that the tree diagram in Figure 1-1 does not indicate any functional categories of traditional clause elements. Furthermore, this example illustrates that the concept of transitivity obviously does not play a major role at the level of syntax in the Minimalist Program. The categorisation of verbs, say into intransitive and transitive ones (with the usual sub-classifcations), is not a matter of syntax in the Minimalist Program, which thus keeps to the generative tradition of viewing syntax as autonomous (and context-free). Rather, transitivity of verbs continues to be an issue to be tackled at the level of lexicon. In Figure 1-1 it is thus the lexical ‘complement feature’ [Obj] of the verb envy (referring to the necessity of an object) that comes into operation.22 22
Radford et al. (1999: 302) note that “complement features ... determine the range of complements which they [words] can or can’t take”. Of course envy may also be used as a ditransitive verb. Then, other complement features would be taken to come into operation.
18
Chapter 1
It is quite obvious that the minimalist approach poses a serious problem for the description of ditransitive verbs and their relation to both obligatorily required objects: the generative axiom that syntax is binary-branching does not allow for an analysis in which the ditransitive verb is equivalently associated with the direct object and the indirect object. If this axiom is to be upheld, one is forced to superordinate one object to the other object in the hierachy of constituents, which would imply that one object is in closer proximity to the verb than the other one. This not only refers to the tree diagram, but also to the underlying model of competence since such tree diagrams are believed to represent physically existent aspects of language-psychological reality: “The grammar represented in the mind is a real object” (Chomsky, 1980: 120); its “constructs and principles can properly be ‘termed mental’” (Chomsky, 2000: 168). The proximity to the verb is of pivotal importance to generativists because in the current ‘checking theory’ it is assumed that case assignment is only possible if the verb and the object (to be case-marked by the verb) are within one constituent, that is, if they display a local specifier-head relation. In this context, the superordinate node of the constituent is considered an invincible ‘barrier’.23 How then are sentences analysed in the Minimalist Program that imply ditransitive verbs and, thus, ditransitive complementation? Larson (1988), who introduces the concept of ‘vp-shells’, has been among the first generativists to delve more deeply into the ‘double object construction’ and its variation. Thus, vp-shells are often also called ‘Larsonian shells’ (cf. Culicover 1997: 364ff.), to which I now turn.24 Figure 1-2, which is taken from Culicover (1997), gives the design of the generative tree diagram of a syntactic structure in which a ditransitive verb is complemented with an indirect and a direct object.25 The problem now is that the verb is in the ‘wrong’ position, as it were. Following generative tenets, case assignment with regard to the indirect object is only possible if the verb is positioned immediately to its left (an assumption which mirrors a major aspect of the older concept of government). Thus, the verb has to be moved leftward, which is done by hypothesising an additional verb phrase (VP) on top of the given tree 23
Note that the programmatic title Barriers of Chomsky’s (1986) book refers to this very assumption. However, in some cases an ‘external checking’ is possible from the generative point of view, but continues to be seen as the exception rather than the rule. Accordingly, such verbs are called ‘exceptional case marking’ (ECM) verbs (Culicover, 1997: 31). Yet, there is no empirical evidence whatsoever for the exceptional status that is ascribed to ECM verbs. 24 It is not necessary here to go into detail about all the modifications and refinements of the Larsonian model that have been offered, for example, by den Dikken (1995: 109ff.) and Brandt (2001). 25 The tree diagram in Figure 1-2 goes back to the ‘c-command’ rule (with ‘c’ standing for ‘constituent’): “A node X c-commands another node Y if the mother of X dominates Y, and X and Y are disconnected (X and Y are disconnected if X ≠ Y and neither dominates the other)” (Radford, 1997a: 112). With regard to the double object construction, this rule leads to the analysis given in Figure 1-2 since – as Culicover (1997: 368) states – “the indirect object must c-command the direct object”.
Ditransitive Verbs in Previous Research and in the Present Work
19
diagram, cf. Figure 1-3. The ditransitive verb, which in Figure 1-3 is give, can then be raised to the empty V-position of the outer ‘vp-shell’.
Figure 1-2:
The starting-point for a Larsonian analysis of the double object construction (Culicover, 1997: 368)
Figure 1-3:
The Larsonian vp-shell analysis (Culicover, 1997: 369)
Broadly, the analysis offered in Figure 1-3 is based on the assumption that there is an outer vp-shell to which the verb moves from its original position because it is said to be merged with a hypothetical ‘light verb’ in the vp-shell (cf. Culicover, 1997: 419 and Radford, 1997b: 201). This light verb is believed to be in the empty V-slot [e] and to have a ‘light’ meaning along the lines of “an abstract causative light verb ∅ – i.e. a null verb with much the same causative interpretation as a verb like make” (Radford 1997b: 201). As for the sentence She gave her a doll, the theory of light verbs in vp-shells results in the analysis given in Figure 1-4. In the tree diagram in Figure 1-4, the ditransitive verb gave is in the wrong position initially (i.e. to the right of the indirect object). It is moved leftward (and simultaneously upward) so that it merges with the hypothetical light verb made in
20
Chapter 1
IP D
I' I [e]
VP Spec [e]
V' V [made]
VP V'
DPio V
DPobj
gave She Figure 1-4:
gave
her
a doll
A vp-shell analysis of She gave her a doll
a newly introduced vp-shell. This light verb has a causative interpretation in the sense of ‘someone/something makes her give...’, which is said to be the semantic contribution of the outer vp-shell to the meaning of the sentence. In principle, this very vp-shell is needed because of the additional V-slot it provides (initially filled with the light verb made) and to which the ditransitive verb gave can be moved: this movement makes it possible, then, to map the tree diagram onto the sentence She gave her a doll. However, for theory-internal reasons it is also considered necessary to move the indirect object leftward (and upward), so that the diagram in Figure 1-4 does not yet provide the complete analysis according to the Larsonian-shell hypothesis. At this stage, Figure 1-4 is just intended to visualise the movement of the ditransitive verb and how it is merged with a light verb. To recapitulate, a solution to the initially wrong position of the verb is thus to draw on abstract light verbs in hypothetical vp-shells. With regard to ditransitive complementation with the indirect object realised as a prepositional phrase, this solution results in a somewhat less complex tree diagram and is therefore visualised first here. Figure 1-5, taken over from Culicover (1997), refers to the verb phrase send a letter to Mary. The corresponding double object construction with both objects realised as noun phrases (i.e. the verb phrase send Mary a letter) is analysed similarly. However, as Figure 1-6, also taken over from Culicover (1997), shows, it is considered necessary to also move the indirect object Mary to the DP-slot of a superordinated constituent in order to meet the generative criteria for case assignment. Figure 1-6 thus visualises the raising of both the ditransitive verb and the indirect object.
Ditransitive Verbs in Previous Research and in the Present Work
21
Figure 1-5:
An analysis of the verb phrase send a letter to Mary (Culicover, 1997: 370)
Figure 1-6:
An analysis of the verb phrase send Mary a letter (Culicover 1997, 370)
The analysis of the corresponding passive constructions requires even more positional shifts of the object for it to become the subject because many more vpshells are now necessary between the original position of the object and the target position of the subject. For the purpose of the present study, the general question arises whether and to what extent the concept of ditransitivity can be integrated into the generative model of light verbs and vp-shells. Is the actual verb (in the inner vp-
22
Chapter 1
shell) ditransitive, or is the hypothetical verb in the outer vp-shell ditransitive, or is ditransitivity inherent in and/or conditioned by both verbs? In a wider setting, one has to wonder whether ditransitivity can be reconciled with strictly binarybranching generative models at all. In particular, the fact that the ditransitive verb equally requires a direct and an indirect object is largely at odds with the generative assumption that case assignment only applies to one object which, then, is more closely related to the verb than the other object. The conclusion must be drawn that the analysis of both objects as mandatory clause elements cannot be adequately mapped onto a generative tree diagram. As already pointed out by Dixon (1965: 169), “insistence on binary splits will tend to obscure the correlations originally recognised, and unnecessarily reduce the faithfulness of a description without significantly adding to its simplicity.” However, this is not so much a problem of graphic representation but of the underlying language theory since tree diagrams are believed to represent the cognitive reality of language in generative grammar. A further theoretical problem that generative grammar poses for a viable description and analysis of ditransitive verbs is the fact that, for the most part, generative grammarians deliberately abstain from taking into consideration functional categories (such as direct and indirect object), let alone semantic roles of the functional categories, but confine themselves to the level of formal realisations at phrase level (and their ‘thematic roles’). In the Comprehensive Grammar, ditransitive complementation is primarily defined in terms of the clause pattern SVOO, that is, by referring to the functional categories that are required by the verb and that correlate with a range of different formal realisations (see section 1.2.1). In generative grammar, on the other hand, objects and all other functional categories are seen as “universal syntactically primitive elements of grammar” (Bresnan 1982: 283).26 They are not systematically applied to syntactic analysis and are thus not represented in the corresponding tree diagrams because a careful consideration of the level of functional categories runs counter to the basic generative view that syntax should be described by means of a strictly formal set of rules.27 26
This has not changed in the Minimalist Program. It is telling that Radford (1997a) uses the term ‘object’ only very sporadically. Furthermore, he explicitly points out in the context of ditransitive verb complementation that ‘object’ is a concept associated with ‘traditional grammar’: “Since complement DPs are referred to as objects in traditional grammar, structures like (35) [They will get the teacher a present, Could you pass me the salt, James Bond showed her his credentials] are said to be instances of the double-object construction” (Radford, 1997a: 377). 27 Note, however, that in Lexical-Functional Grammar (LFG), grammatical relations are no longer treated as primitives (cf. van Valin, 2001: 191). What is taken over from the established generative tradition, though, is the exclusive focus on isolated sentences and on syntactic constraints and features (with a significant role ascribed to the lexicon) as well as a particular interest in the typological variety of syntactic structures across languages. Thus, the ‘functionalism’ of LFG does not refer to the context-dependent discourse function of syntax, but to the consideration of semantic restrictions in a context-independent model of syntax. For an overview of LFG see Bresnan (1982, 2000).
Ditransitive Verbs in Previous Research and in the Present Work
23
The suggestion that ditransitive verbs (and also complex-transitive verbs) are merged with abstract light verbs in vp-shells has found wide-spread acceptance in generative grammar and can be found in standard introductions to the Minimalist Program such as Radford (1997a, 1997b). In my view, however, the whole chain of argumentation is based on a very shaky foundation. In general, since performance data are considered irrelevant to the understanding of the internalised competence, it is quite clear that the Minimalist Program as such has little to offer to descriptive linguists who are predominantly interested in actual language use, as for example the use of ditransitive verbs in real contexts by real speakers. In particular, the vp-shell argument outlined above brings to light the fact that generative models of competence can never be verified or falsified empirically as long as they are not based on large amounts of observable and testable data but only on very few invented and decontextualised sentences. But not only the lack of linguistic data is a problem; even more serious is the lack of neurobiological evidence, as generative grammar is in fact an attempt to develop a model of language cognition, of language in the mind, i.e. ‘I-language’. This being so, it is somewhat puzzling to see the rise of the vp-shell hypothesis without any cognitive data supporting this assumption. As things stand, I do not see any compelling reason for adopting the highly speculative and merely intuition-based vp-shell hypothesis because there nothing in the data to make its adoption plausible. Rather, it appears to me as though the vp-shell and the light verb in it have been suggested largely in order to overcome an analytical problem caused by the generative model itself. That it is virtually impossible to verify or falsify any generative account on the basis of observable data is also corroborated by the generative notion of ‘dative shift’ or ‘dative movement’. In early transformational grammar, ‘dative movement’ served as a classic example (besides, say, the active-passive alternation) of a transformational rule by means of which two surface structures were shown to be derived from one and the same deep structure. Baker (1978), for example, analyses the transformation between examples (4) and (5). (4) (5)
Alice sent a book to Alfred Alice sent Alfred a book
Baker (1978) formulates the transformational rule to be applied here as follows: Dative Movement X – V – NP – to – NP 1 2 3 4 5 => 1, 2 + 5, 3, 0, 0 (Optional)
28
(Baker, 1978: 248)28
In the formula for the transformational rule, the first zero refers to the omission of the preposition to. The second zero indicates that the old position of ‘5’ remains empty because the NP in ‘5’ is now placed before the NP in ‘3’.
24
Chapter 1
He does not ignore the fact that some verbs, such as report, convey and say, do not allow for dative movement. This problem can be solved, Baker (1978: 249) writes, “if we include the rule feature <–Dative Movement> in the lexical entries of the verbs report, convey and say.” The important point here is that at the level of syntax, a general transformational rule of dative movement continues to be postulated while exceptional cases are dealt with in the lexicon. One may approve of this transformational rule or not: what is surprising is the fact that other generative models offer completely different accounts. Ouhalla (1994: 145), for example, discusses a very similar pair of examples:29 (6)
Mary gave the book to John Mary I [VP gave [NP the book] [PP to John]]
(7)
Mary gave John the book Mary I [VP gave [NP John] [NP the book]]
Ouhalla (1994) holds the view that the syntactic changes involved here are not a matter of transformation because they cannot be captured by a generally applicable transformational rule: There is evidence which suggests that the pattern [V NP NP] shown in (64) [here 10b], called Dative Shift, is unlikely to be derived by a transformational rule from an underlying structure with the order [V NP PP]. Although a substantial number of verbs which select two internal arguments, e.g. give, send, buy, allow the Dative Shift pattern, there are verbs which do not. For example, although the verb donate is close in meaning to the verb give, it does not tolerate the Dative Shift pattern ... . Transformational rules are usually not sensitive to individual lexical items, given that they operate on (classes of) categories. (Ouhalla, 1994: 145f.) Specifically, while Baker (1978) suggests a transformational rule that is responsible for the so-called dative movement, Ouhalla (1994) rejects such a rule on grounds of its inapplicability to all ditransitive verbs. In my view, the two fundamentally different positions highlight a methodological and a related theoretical problem, both of which are typical, I feel, of mainstream generativism in general. The methodological problem refers to the fact that there is no empirical and/or testable evidence for either position.30 Why does Baker speak of a transform29
Note that both Baker (1978) and Ouhalla (1994) still use the term ‘noun phrase’ (NP) which has recently been replaced with ‘determiner phrase’ (DP). 30 Consider in this context also van Valin’s (2001: 191) comparison of different generative approaches to dative shift and active-passive alternation. He points out that both phenomena have been described in terms of lexical rules (e.g. in Lexical-Functional Grammar) and in terms of syntactic rules (e.g. in Relational Grammar).
Ditransitive Verbs in Previous Research and in the Present Work
25
ational rule at work here and try to come to terms with obvious exceptions at the level of the lexicon? Is it because the exceptions are exceptions in quantitative terms? This appears doubtful since frequency in text has never been considered relevant in the generative paradigm. In fact, Givón (1984) calls into question the usefulness of the concept of dative shift because the syntactic arrangement with a seemingly shifted dative turns out to be the unmarked case in actual language use: It is normally taken for granted that the promotion of DAT/BEN [dative/ benefactive] objects to DO [direct object] via Dative Shifting is an ‘optional’, ‘stylistic’ device in English and other languages (Indonesian). However, a careful study of frequency distributions in live text reveals that in such unsolicited discourse the overwhelming bulk of DAT/BEN objects appear as DO rather than IO. (Givón, 1984: 155) In the light of these findings, it may well be argued that – contrary to the dativeshifting hypothesis – it is (7) rather than (6) which is more basic. On the other hand, why does Ouhalla reject the applicability of a transformational rule to the dative shift altogether? In his line of argumentation it seems as if he suggests that one cannot speak of transformation whenever the postulated requirements of transformational rules are not met. However, he takes for granted that transformational rules as defined by generativists do exist in the psychological reality of language. This leads us to the theoretical problem mentioned above. Many concepts and categories, e.g. transformation, which have been posited in generative grammar on the basis of invented and isolated sentences (and the intuition-based judgment on their grammatical well-formedness) prove irrelevant when it comes to the careful analysis of large amounts of authentic data. In this context, Gross (1979) is perfectly to the point: ... when one possesses an extensive picture of a language, i.e. a categorization of the great bulk of the lexical elements and their local constraints, one sees that the formal notion of transformation, as promulgated by GG [generative grammar], is of marginal importance. GG bears only on insignificant and arbitrary parts of the materials. Moreover, it has never developed the means to verify its limitations. (Gross, 1979: 873) Other concepts in generative grammar as well, for example vp-shells and light verbs, represent mere speculation about linguistic competence: their existence, let alone their suitability for grammatical description, is a matter of belief rather than of linguistic evidence and descriptive plausibility. Note that in the Minimalist Program, vp-shells and light verbs are also drawn on in order to account for the systematic correspondence between ditransitive variants, i.e. [V NP NP] and [V NP PP]. Chomsky (1995) writes:
26
Chapter 1 V [gave] raises to the empty main verb position of the higher VP shell, yielding John gave a book to Bill. Alternatively, operations similar to those yielding the passive construction could ‘absorb’ the Case of Bill, forcing it to raise to the subjectlike position of a book, which in turn becomes an adjunct, yielding John gave Bill a book. (Chomsky, 1995: 62f.)
The big imponderable is that, unfortunately, Chomsky does not explain how case can in fact be ‘absorbed’. Furthermore, it remains a mystery to me why a book in John gave Bill a book should be considered an adjunct. It is also telling that in the recent Cambridge Grammar of the English Language, Huddleston and Pullum (2002: 1038) refrain from applying the Larsonian analysis and make use of a ternary analysis although there is a clear preference for binary analyses throughout the grammar (which is strongly influenced by generative concepts and rigidly formalist classifications in general; cf. Leech, 2004). In conclusion, the application of the generative framework to the description of ditransitive verbs is fraught with problems, as the examples discussed in this section reveal. Anagnostopoulous’s (2003) large-scale typologically-oriented study of ditransitive verbs does not solve any of these problems since she does not critically review the theory-internal assumptions and does not call into question existing categories and labels (e.g. vp-shells) which are neither derived from authentic data nor corroborated by any other kind of empirical evidence. As Trotta (2000) points out: ... classifications and labels can be very misleading and create a kind of virtual reality which may blind the researcher and prevent them from open-mindedly studying the one which actually exists. (Trotta, 2000: 6) My impression is that this is precisely what has been happening in mainstream generative grammar for quite some time. 1.2.3 Valency theory Valency theory has evolved from Tesnière’s (1953) dependency theory. In this framework, verbs (but also nouns and adjectives) are seen as items which to a considerable extent govern their syntactic environment because of the valencies that are inherent in these items. In his initial approach to verb valency, Tesnière (1959) takes into account quantitative valencies, i.e. the different number of clause elements required by different verbs. Within this framework, ditransitive verbs such as give are labelled ‘trivalent’ since they must be complemented with a subject and two objects.31 Furthermore, verbs are analysed in terms of their 31
I will, however, continue to speak of ‘ditransitive’ verbs for two reasons. First, this term has found more general acceptance for verbs requiring two objects. Second, the term
Ditransitive Verbs in Previous Research and in the Present Work
27
qualitative valency, i.e. in regard to the kinds of obligatory clause elements (e.g. noun phrase, prepositional phrase, that-clause, to-infinitive).32 Apart from the quantitative and qualitative valency of verbs, particular attention has been paid to the question whether and to what extent clause elements are in fact obligatory (cf. e.g. Allerton, 1982: 61ff. and Klotz, 2000: 11ff.). In bringing together previous attempts to systematise different kinds of valency, Klotz (2000: 14) suggests a gradient with truly obligatory clause elements placed at one end of the gradient (‘obligatorische Ergänzungen’) and entirely optional elements at the other end (‘freie Angaben’). In between obligatory and optional elements we find different kinds of elements that may be omitted for various reasons (‘fakultative Ergänzungen’). For example, clause elements may be deleted (and the corresponding valency of the verb thus suppressed) because they “can be assumed to be contextually evident to the listener” (Allerton, 1982: 69): such elements would be contextually optional (‘kontextuell-fakultative Ergänzungen’). Clause elements may also be omitted because they are easily inferrable from world knowledge (‘stereotyp-fakultative Ergänzungen’). As a matter of fact, there are many problems of demarcation involved. For example, Klotz (2000: 13) mentions instrumental with-PPs, which are principally optional elements (e.g. He destroyed his computer [with a virus-infected disk]) but may become the notional subject of the sentence, which is more typical of obligatory clause elements (e.g. A virus-infected disk destroyed the computer). However, it is not necessary to go into details about such problems of demarcation since ditransitive verbs, which are of interest here, have to be complemented with two obligatory objects. It has to be admitted, though, that many ditransitive verbs can also be used in the monotransitive clause pattern with one object alone (see section 1.2.1). But the in-depth analysis of corpus data in chapter 3 will provide ample testimony of the fact that the deleted object is contextually reconstructible in the vast majority of cases. The working premise at this stage is that neither the indirect nor the direct object (let alone the subject) is to be considered an entirely optional clause element. A particularly important aspect of recent valency theory is the distinction of syntactic and semantic valency. The concept of syntactic valency refers to the range of syntactic structures which are permitted with specific verbs at the level of formal realisations of obligatory clause elements. In a similar vein to the Comprehensive Grammar (see section 1.2.1), Klotz (2000: 14) views Jenny in John gave Jenny a bunch of roses and to Jenny in John gave a bunch of roses to Jenny as variations of one and the same valency slot since the noun phrase Jenny and the prepositional phrase to Jenny are co-referential. From the point of view of valency theory, the two sentences are thus considered equivalent. This example ‘trivalent’ would also refer to complex-transitive verbs in the clause patterns SVOC and SVOA. 32 The necessarily brief overview in this section owes a large debt to previous discussions of verb valency, particularly by Emons (1978), Allerton (1982) and Klotz (2000). Note that Klotz (2000) also provides a corpus-based analysis of some ditransitive verbs by drawing on the Bank of English Corpus. Some of his results and conclusions obtained from corpus data will thus be discussed at a later stage (see chapter 3).
28
Chapter 1
illustrates the fact that in valency theory, too, analogies are drawn between different formal realisations at the level of functional categories. This, of course, implies that in both sentences give is regarded as a ditransitive verb although the clause pattern SVOO is only apparent in the sentence John gave Jenny a bunch of roses, whereas in John gave a bunch of roses to Jenny one might hypothesise a complex-transitive verb give in the clause pattern SVOA (as suggested by Standop (2000), see section 1.2.1). The concept of semantic valency refers to the meaning-side of the clause elements required by the meaning of the verb. In principle, the semantic level of valency is more basic than the syntactic one because syntactic valency is conditioned by the semantic valency of a given verb. With regard to the ditransitive verb give, Klotz (2000: 16) points out that it is the inherent semantic structure of the verb, which he paraphrases as “A causes B to have C”, that leads to the syntactic trivalence of give. Thus, syntactic valency is a secondary phenomenon: syntactic elements are required by the verb because of the meaning of the verb. Also, the syntactic elements can be viewed as formal realisations of underlying ‘semantic roles’. This is in line with Fillmore’s (1968: 24) set of ‘case roles’, which has exerted an enormous influence on many linguistic studies and grammars such as the Comprehensive Grammar with its ‘semantic roles’ (see section 1.2.1). Accordingly, the concept of semantic roles in the argument structure (underlying ditransitive complementation) will play a major part in the definition of ditransitivity in the present study (see section 1.3.1).33 Valency theory has much in common with the structuralist approach of the Quirk-grammars. Although Quirk et al. (1985) use the term ‘valency’ as such only once, the important aspects and implications of the concept of valency are covered by their preferred term ‘complementation’ (see section 1.2.1): the requirement of two objects in the complementation of ditransitive verbs refers to quantitative valency, the range of formal realisations of clause elements that a specific ditransitive verb allows for refers to qualitative valency, and the distinction between clause elements as syntactic categories and the semantic roles that they fulfil mirrors – at least, generally – the distinction between syntactic and semantic valency. Proponents of valency theory have always seen syntax and semantics as closely related. As has been noted before (see section 1.2.1), this interrelation is not made very explicit in the Comprehensive Grammar with regard to ditransitive verbs in particular. The following section is thus devoted to some major approaches in functional grammar that have investigated more closely the interrelation of ditransitive syntax and semantics.
33
The concept of semantic roles plays an important part in virtually all functional grammars (see section 1.2.4). The construction grammar approach goes back to Fillmore’s (1968) seminal article on case roles and will be discussed more closely in section 1.2.7.
Ditransitive Verbs in Previous Research and in the Present Work
29
1.2.4 Functional grammar and semantico-syntactical approaches ‘Functional grammar’ is used here as a generic term to describe a wide range of linguistic approaches to grammatical relations.34 Although many theories and models of grammars have been given the label ‘functional’ or ‘functionalist’, it is important to note that “[t]he term function is not unambiguous” (Daneš, 1987: 4). It is thus necessary to briefly sketch out major aspects of the inherent ambiguity of the term ‘functional’ and to clarify how the term will be used in the present study. Daneš (1987) distinguishes five different types of ‘functional’ approaches to language that can be found both in linguistics in general and in Prague-school linguistics in particular: 1. the methodological device to start from the common needs of communication of expression (= from functions) and to ask by what means these needs are satisfied in the analysed language; a comparison of different languages without regard to their genetic relations is possible chiefly on the basis of the common needs of communication and expression; 2. external functions of language (and/or utterances); 3. functions of the units of a language system; 4. some other cases of the use of the attribute functional [e.g. the ‘functional load’ of minimal pairs in phonology]; 5. functional explanation of language development. (Daneš, 1987: 9) The use of ‘functional’ for methodological (e.g. contrastive and typological), diachronic and other purposes (as for example in the context of ‘functional load’ of linguistic elements) – i.e. definitions 1, 4 and 5 – are of minor relevance to the present study. Of particular importance are definitions 2 and 3: they embrace language-external and language-internal functions respectively. External functions of language – and, thus, of grammar – refer to communicative (or poetic or intellectual) goals that speakers want to achieve by using language in a given discourse context.35 In this category, grammatical structures and devices are viewed as fulfilling communicative and, in the widest sense, pragmatic functions. The use of grammatical structures can be explained, for example, by means of pragmatic principles (such as end-focus and end-weight) as well as considerations of given and new information (or theme and rheme) that may explain why in given contexts language users prefer particular grammatical choices. Internal functions, on the other hand, include the functions that linguistic units of a particular level fulfil at another linguistic level. This category thus picks up on 34
Thus, ‘functional grammar’ in this context is not restricted to – but is meant to include – Dik’s (1978, 1989, 1997) theory of functional grammar. 35 This aspect is closely related to Jakobson’s (1960: 357) discussion of the ‘six basic functions of verbal communication’. See also the discussion offered by König (1977: 1ff.).
30
Chapter 1
Vachek’s (1976: 72) concept of ‘functional complementation’: “the components of a lower level are functionally complemented into a unit of the neighbouring higher level.” At clause level, various formal realisations (e.g. noun phrase, thatclause, to-infinitive) can be taken to be functionally complemented into ‘functional categories’ at the neighbouring higher level, i.e. clause elements (e.g. direct object).36 A third domain of the term ‘functional’, which sits somewhat uneasily on the boundary between language-external and language-internal functions, is the exploration of meaning in language. Sgall et al.’s (1973) model of a ‘functional generative grammar’ is a good example of many frameworks in which ‘functional’ is used in the sense of ‘semantic’: “it appears that the characteristic ‘functional’ in the approach of Sgall’s group only underlies the semantic character of their generative conception” (Daneš, 1987: 22). At the level of ditransitive verb complementation, this meaning of ‘functional’ is closely related to the languageexternal, communicative/pragmatic dimension of ‘functional’ (see above).37 The three relevant dimensions of the notion of ‘function(al)’ are summarised in the upper part of Figure 1-7: (a) in terms of ‘functional categories’; (b) in terms of ‘semantic’; (c) in terms of ‘communicative/pragmatic’. function / functional
(a) function of a linguistic unit on another linguistic level (= ‘internal’ function) => ‘functional category’ (clause elements) Figure 1-7:
(b) function as meaning of linguistic form (= semantics)
(c) communicative/ pragmatic function in discourse (= ‘external’ function)
=> ‘function(al)’ (semantic and pragmatic)
Dimensions of ‘function’/‘functional’
As shown in the lower part of Figure 1-7, the term ‘function(al)’ will be used in the present study for aspects of semantic and language-external (pragmatic) functions of linguistic forms. This is in line with Dik’s (1987) definition of ‘functional’, which embraces semantic and pragmatic aspects:
36
Note that the postulation of language-internal functions is peculiar to Prague-school linguistics and rejected by most American functionalists (cf. Newmeyer, 2001: 113f.). 37 For example, specific realisations (that-clause vs. noun phrase) of a clause element (e.g. direct object) required by a ditransitive verb (e.g. tell) fulfil at the same time semantic and pragmatic functions (see section 3.1.2.2).
Ditransitive Verbs in Previous Research and in the Present Work
31
In the Functional Paradigm, syntax cannot be regarded as autonomous with respect to semantics. Rather, the very essence of syntax is that it provides the means for creating meaningful expressions. And the system of language cannot be regarded as autonomous with regard to pragmatics. Rather, the very essence of language is that it must function properly and effectively in verbal interaction. (Dik, 1987: 82) The only level at which the present study of ditransitive verbs will have to take into consideration language-internal functions of linguistic forms is at the level of clause elements, i.e. subject, direct object, indirect object. In the present study, I will refer to these clause elements as ‘functional categories’. What all grammars that are functional in the sense of Figure 1-7 have in common is the central assumption that grammatical forms are dependent on the functions they fulfil and that, consequently, grammatical forms can be explained in functional terms. In a weaker sense, the Quirk-grammars can also be subsumed into the category of functional grammars. In the present section, however, it is my intention to look more closely at functional grammars in the fullest sense, namely at grammatical descriptions in which the linguistic connection between syntax and semantics provides the main guideline for grammatical analysis. Major examples of what I would like to call ‘semantico-syntactical approaches’ to grammar are Jackson’s (1990) Grammar and Meaning, Dixon’s (1991) New Approach to English Grammar on Semantic Principles and Halliday’s (1994) Functional Grammar. In the following, I will explore in particular what the three studies have to offer with regard to ditransitive verbs.38 Jackson’s (1990: 159) semantic interpretation of the clause pattern SVOO has already been mentioned in section 1.2.1. By scrutinising the sentence Bill offered Jane a chocolate on the bus last night, Jackson (1990) describes why the verb requires specific semantic roles to be filled:39 ... the participants in the drama of ‘offering’ are Bill, Jane and a chocolate. The drama of ‘offering’ has the cast of ‘offerer’, ‘offeree’ and ‘thing offered’; they have the semantic roles of AGENTIVE, RECIPIENT and AFFECTED respectively. ... The participants are more or less obligatory elements in a proposition. A particular
38
Horie and Comrie (2000: 3f.) point out that the borders between functional grammar and typological studies as well as cognitive linguistics are fuzzy. The combination of the three perspectives (i.e. language functions, language typology and language cognition) is at the basis of the chain of argumentation in very influential monographs such as Comrie (1989), Croft (1990) and Givón (1995). Note also that it is beyond the scope of the present study to consider functional grammars in their entirety. Other important works in this field include, for example, Green (1974), Givón (1979), Cattell (1984) and Wierzbicka (1988). 39 The term ‘drama’ is used to denote the abstract features of the underlying proposition: “A proposition ... may be seen as representing a kind of drama” (Jackson, 1990: 46).
32
Chapter 1 situation type, represented by a verb with a particular meaning, expects a certain cast of participants ... . (Jackson, 1990: 46)
While the participants are more or less obligatory, the circumstances – i.e. the adverbials on the bus and last night in the example at hand – are dispensable: “Circumstances, on the other hand, cannot be generally said to be necessary for the completeness of a proposition” (Jackson, 1990: 47). This certainly resembles the distinction of obligatory and optional clause elements in valency theory (cf. section 1.2.3).40 On the strength of his focus on primarily semantic relations and dependencies, Jackson (1990) is now able to explain much more plausibly the possible omission of contextually reconstructible participants, e.g. in the sentence I’ve brought you lovely, lovely presents (i.e. sentence [K28:82] taken from the LOB Corpus) and other bring-sentences: ... a proposition referring to the situation type bring will contain three elements in addition to the ACTION: (1) AGENTIVE, (2) AFFECTED, (3) LOCATIVE (SOURCE/GOAL) or RECIPIENT. The third element may be omitted under certain contextual conditions, viz if it is ‘here’. We may say therefore that these elements are semantically obligatory with bring, which determines their presence and on which they are dependent. (Jackson, 1990: 139) It should be noted in passing that this interpretation could also explicate the example of She gave a doll which Standop (2000: 223) mentions (see section 1.2.1). In this example as well, give remains a ditransitive verb. That is, the underlying situation type is of the usual GIVING type. The recipient is simply omitted because it is inferrable from the context or because a specification of the recipient is irrelevant to the context. Jackson (1990) also refers to the similar case of We’ll give a nice one and argues that the seemingly incomplete syntax can be explained if the actual context is considered so that the meaning (and semantic valency) of the verb give is left intact: [99] We’ll give a nice one A first reaction might be to declare [99] ungrammatical, and to ask ‘Who’ll be given a nice one?’ But if we imagine [99] said in the context of donating prizes for a competition or items for a charity auction, then it begins to sound more grammatical. (Jackson, 1990: 161)
40
Note that Jackson also sees the ‘verb with a particular meaning’ as the item which governs both semantic roles and syntactic structures in the sentence. A comprehensive list of semantic groupings of verbs is provided by Levin (1993: 111-276).
Ditransitive Verbs in Previous Research and in the Present Work
33
In a similar vein, the direct object may also be deleted for contextual reasons, as for example in the following case: 41 [100] We give to Oxfam An appropriate context for [100] might be as a reason for not putting anything in the collecting tin of an alternative charity. (Jackson, 1990: 161) Basically, this functional analysis is an operationalisation of Matthews’s (1981: 125) concept of ‘latent’ objects – that is, objects that are semantically obligatory, but that are dropped syntactically because they are contextually reconstructible. The important thing here is that in such cases there may be an incomplete sentence but not an incomplete utterance due to the latency of the object not explicitly expressed (cf. Matthews 1981: 38ff.). Whether a sentence, say with a ditransitive verb, is in fact an incomplete sentence or an incomplete utterance can be clarified only if the context is looked at. This is where the corpus as a large and representative sample of authentic language use comes into play. By considering the context of language use, corpus analyses may help identify context-dependent factors which may lead to a deletion of otherwise obligatory clause elements. In the light of Jackson’s aforementioned examples (which are taken from the LOB Corpus), it proves necessary to clearly distinguish between syntactically and semantically obligatory elements (or, between syntactic and semantic valency) because the omitted clause element remains semantically obligatory (but is not necessitated syntactically). It is quite obvious that descriptive and structuralist grammars, which are predominantly oriented towards the norm of syntactically well-formed (and, thus, complete) sentences, make it difficult to put such a distinction into practice. It goes without saying that in generative grammar, which is based on the assumption that syntax is autonomous (and, in principle, independent of semantics), it is even more difficult to allow for syntactically incomplete but at the same time semantically complete sentences. Functional grammar, on the other hand, is a particularly useful point of reference since at its heart lies the principal assumption that grammar and meaning are virtually inseparable: “grammar and meaning combine to enable the multitude of communication functions that we call upon language to serve” (Jackson, 1990: 256). From a functionalist point of view, it is reasonable to assume that ditransitive complementation represents not only a syntactic but also a semantic phenomenon. More specifically, ditransitive verbs are semantically similar in that the underlying situation types, including the number and kind of participants involved, are similar. It is due to these semantic commonalities that ditransitive verbs are associated with the ditransitive complementation syntactically. To a certain extent, ditransitivity thus provides an example of iconicity between syntax
41
This makes perfect sense since in corresponding passive constructions the by-agent may also be readily omitted.
34
Chapter 1
and semantics: “language has a general iconic tendency whereby semantic sameness is reflected also by formal sameness” (Anttila 1972: 89).42 The interrelation of syntax and semantics is emphasised to an even larger extent by Dixon (1991: 6), who starts off by classifying nouns, adjectives and verbs into ‘semantic types’, defined as “large classes [of words] that have common meaning components”, in order to identify similarities in grammatical structures that are associated with words of the same semantic type. For the purpose of the present work, what Dixon (1991) has to say about the semantic type of ditransitive verbs is of particular importance: A set of verbs is grouped together as one semantic type partly because they require the same set of participant roles. All GIVING verbs require a Donor, a Gift and a Recipient, as in John gave a bouquet to Mary, Jane lent the Saab to Bill ... . (Dixon, 1991: 9) In a similar vein to Jackson’s approach, Dixon thus assumes that the meaning of the ditransitive verb triggers off the typical ditransitive situation type (or ‘semantic type’). Dixon’s methodology, though, is slightly different from Jackson’s in that he first classifies an amazingly great number of verbs into semantic types and then analyses the differences in verb complementation on semantic principles.43 That is, Dixon proceeds from the level of lexical meaning to syntax and and its semantics. This procedure leads to a variety of semantic types and subtypes of ditransitive verbs. Concerning lexical verbs, Dixon (1991) distinguishes between ‘Primary-A verbs’ and ‘Primary-B verbs’. Primary-A verbs choose an argument structure “with roles filled by NPs that have CONCRETE Heads” (Dixon, 1991: 94).44 On the other hand, Primary-B verbs “can also, unlike Primary-A, have a complement clause as alternative to an NP in one syntactic relation (e.g. John saw that Mary had won)” (Dixon, 1991: 124). Within Primary-A verbs, Dixon (1991: 94ff.) distinguishes between the following semantic types: (a) MOTION and REST (e.g. run, kneel); (b) AFFECT (e.g. hit, cut); (c) GIVING (e.g. give, donate); (d) CORPOREAL (e.g. eat, laugh); (e) WEATHER (e.g. rain, snow); (f) some further types such as COMPETITION and OBEYING. 42 Haiman (1980: 515) uses the term ‘motivation’ to capture such correspondences: “the structure of language directly reflects some aspects of the structure of reality.” 43 On the other hand, Jackson’s (1990) methodology is more oriented towards grammatical structures such as clause patterns. 44 ‘CONCRETE heads’ are nouns with a “CONCRETE reference, e.g. youth, horse, foot, piece, grass, star, fire, hill, city, table” (Dixon, 1991: 76).
Ditransitive Verbs in Previous Research and in the Present Work
35
The vast majority of ditransitive verbs occur in the GIVING type: “Verbs of this type involve three semantic roles – a DONOR transfers possession of some GIFT to a RECIPIENT” (Dixon, 1991: 113). Even though Dixon himself does not use the term ‘ditransitive’ as such, he implicitly regards all possible formal realisations of the semantic roles involved (in the complementation, say, of the verb give) as representing the same semantic type of GIVING, in this resembling the Comprehensive Grammar. Accordingly, he also groups verbs such as supply with into this semantic type, although they do not occur in the basic form of ditransitive complementation (i.e. with both objects realised as noun phrases). But they are semantically similar to ditransitive verbs proper. Some ditransitive verbs can be found in other semantic subtypes, for example the STAB subtype (e.g. cut) within the semantic type of AFFECT. In this semantic type, the ‘recipient’ is considered ‘benefactive’: examples such as I cut Mary a slice of bread and I knitted Mary a jumper are mentioned by Dixon (1991: 283f.). Also, ditransitive verbs are among the Primary-B verbs (i.e. with clausal complementation), in particular in some subtypes (esp. TELL and FORGIVE with the prototypical examples of tell and forgive) within the semantic type SPEAKING. In conclusion, Dixon thus identifies three general semantic types with which ditransitive verbs are associated: (1) GIVING, (2) AFFECT, (3) SPEAKING. Halliday’s (1994) terminology is differs considerably from Jackson’s and Dixon’s, but the underlying concepts are quite similar: the situation type/ semantic type is labelled ‘process type’ (with each process type correlating with a specific ‘category meaning’), and the semantic roles are called ‘participants’. Ditransitive verbs are thus said to be associated with process types in which the verb requires three partcipants, namely ‘actor’, ‘goal’ and ‘beneficiary’: “Actor is ‘logical subject’ and Goal is ‘logical direct object’, the Beneficiary is ‘logical indirect object’” (Halliday, 1994: 144). As in Dixon’s categorisation, ditransitive verbs occur in three different process types implying the typical ‘beneficiary’: “The Beneficiary is the one to whom or for whom the process is said to take place. It appears in material and verbal processes, and occasionally in relational” (Halliday, 1994: 144). Accordingly, Halliday (1994: 145) offers the following analysis of the sentence I gave my love a ring that has no end: (8)
I
gave
my love
a ring that has no end
Actor
Process: material
Beneficiary: Recipient
Goal
This analysis again throws light on the central role of the verb in functional grammar since it is the verb which determines the process type and, thus, the number and kind of participants involved. Beneficiaries (and, accordingly, ditransitive verbs) not only occur in material processes as in (8), but also in verbal processes (e.g. John told Mary a story) and in some relational processes (e.g. it cost him a pretty penny). Note that the Hallidayan categories correlate to a considerable ex-
36
Chapter 1
tent with Dixon’s aforementioned semantic types of GIVING (~ ‘material process’), SPEAKING (~ ‘verbal process’) and AFFECT (~ ‘relational process’).45 In all three functional approaches to ditransitive verbs that have been reviewed so far, some sort of shared semantic core is ascribed to the ditransitive complementation. Verbs are complemented ditransitively (and are, therefore, ditransitive verbs) because they require the same number and similar kinds of semantic roles. If we adhere to the semantic types suggested by Dixon (resembling, of course, the Hallidayan process types), the general conclusion can be drawn that ditransitive verbs denote concrete, abstract or verbal transfers in the extralinguistic world. This semantic quality of ditransitive verbs and their complementation is visualised in Figure 1-8.
acting entity
transferred entity
ditransitive verb
affected entity
GIVING or ACTING or SPEAKING ditransitive complementation Figure 1-8:
The semantic core of ditransitive complementation.
Figure 1-8 shows the ditransitive verb as the central element, which is associated with a particular configuration of three semantic roles. The three semantic roles and, accordingly, the ditransitive complementation (offering three corresponding syntactic slots) are required by all ditransitive verbs. The semantic type of the transfer inherent in all ditransitive verbs can be of three kinds: GIVING (i.e. a concrete and/or immediate transfer), AFFECT (i.e. an abstract or non-immediate transfer) or SPEAKING (i.e. a transfer of language). The two lines between the acting entity and the transferred entity as well as between the transferred entity and the affected entity reflect the the chronological order of the transfer event in that the transferred entity is transferred from the acting entity to the affected 45 It should be noted that Halliday (1994), unlike the Comprehensive Grammar (cf. Quirk et al. 1985: 1215f.), does not regard to-infinitive clauses as potential direct objects of ditransitive verbs. Accordingly, when analysing the sentence he told Fred to do it, Halliday (1994: 145f.) argues that Fred is not the ‘receiver’ (and, thus, not the ‘beneficiary’) of the verbal process of TELLING, but rather the semantic or notional subject of a separate process denoted by the to-infinitive as such.
Ditransitive Verbs in Previous Research and in the Present Work
37
entity in the course of the underlying event. The dotted line between the acting entity and the affected entity illustrates the fact that the two entities are interrelated indirectly in that the affected entity is affected by the acting entity only via the transfer of the transferred entity.46 Thus, Figure 1-8 visualises the working premise at this stage that the ditransitive complementation represents a specific underlying semantic situation type, which is independent of a particular ditransitive verb but has to be specified by the given verb. Finally, Figure 1-8 indicates that at the level of the situation type underlying the ditransitive complementation, nothing specific is said about the linear order of entities nor about their possible formal realisations. Rather, it provides a rough-and-ready visualisation of the general propositional features of ditransitive complementation which are not yet bound to a specific surface structure. As will be discussed later in this chapter, there is no one-to-one relationship between semantic roles and specific positions in the linear order of elements nor between semantic roles and particular functional categories. 1.2.5 Corpus-based grammar While semantic types, as for example in Dixon’s (1991) sense, are usually the starting-points for functional grammarians, corpus-based grammar is characterised by a fundamentally different procedure, which is largely due to its distinct methodology. In corpus-based grammar, linguists first of all look for formal structures in language use (as represented by large corpora) by using appropriate software programs. In a second step, the data obtained from corpora are analysed quantitatively and qualitatively, including the semantic interpretation of the data and the establishment of semantic categories whenever necessary and possible. The major advantages of corpus-based methodology for descriptive linguistics have already been discussed in detail in previous studies (cf. e.g. Sinclair, 1991; Fillmore, 1992; McEnery and Wilson, 1996; Stubbs, 1996; Tognini-Bonelli, 1996; Biber et al., 1998; Kennedy, 1998; Lenz, 2000; Stubbs, 2001; TogniniBonelli, 2001; Meyer, 2002; Mukherjee, 2002a). It should therefore suffice here to summarise the advantages by quoting Hunston and Francis (2000): The data is authentic; The data is not selected on linguistic grounds; There is a lot of data; The data is systematically organised; The data is not annotated in terms of existing theories. (Hunston and Francis, 2000: 15)
46
It is of some interest that Beermann (2001) correctly points out that the strong link between ditransitivity and transfer events is peculiar to English and cannot be found in German, for example.
38
Chapter 1
In short, the analysis of large and representative corpora by means of modern software facilities makes it possible to describe and analyse linguistic forms and structures rapidly and exhaustively, empirically and reliably, resulting in a testable and frequency-based account of authentic language use. The ‘corpus revolution’ (Crystal, 1995: 438) has led to a focus on actual data and frequencies in text. In this, the relevance of authentic discourse contexts to a functional analysis of syntax cannot be exaggerated. As Mair (1990) points out, syntax and discourse are closely interrelated because the use of specific syntactic structures can be shown to be functionally motivated on grounds of discourse-dependent principles and factors: Syntactic structures ... are not autonomous but closely connected with and partly determined by the conventional structure of discourse, by the desire to express meaning efficiently and, ultimately, by the habits – acquired or genetically programmed – of human perception and cognition. (Mair, 1990: 219) It goes without saying that corpora as large samples of natural discourse provide databases which enable the linguist to investigate the interrelation of syntax and discourse to an extent not possible before. By drawing on corpus-linguistic methodology, the present study is thus an attempt to study syntax in discourse. How corpus-linguistic principles – and problems – translate into the methodology of the present study will be discussed in detail in chapter 2. In this section, I am mainly concerned with previous corpus-based research into ditransitive verbs. The first consistently corpus-based grammar of English is Sinclair’s (1990) Collins COBUILD English Grammar (CCEG), designed as a grammar for learners of English who are interested in actual English usage. The third chapter of this grammar – given the title ‘Making a message’ – is, among other things, about verb complementation. The relevant terminology applied in the CCEG is unusual in two regards. First, transitivity is no longer associated with the verb but seen as a clausal phenomenon: In this grammar, transitivity is explained in relation to the clause as a whole. Many grammars deal with transitivity in terms of the way in which a verb selects objects. However, the verb, while usually central in clause structure, is not necessarily chosen first. (Sinclair, 1990: 137) Second, Sinclair (1990: 172ff.) restricts the term ‘complementation’ to clause patterns implying either a subject complement or an object complement. Hence, he obviously does not distinguish between the concepts of complement (as a specific clause element) and complementation (as all elements required to complete the meaning of the verb). Neither of the two terminologies used in the CCEG will be taken up in the present study. As has been discussed at length in
Ditransitive Verbs in Previous Research and in the Present Work
39
sections 1.2.1, 1.2.3 and 1.2.4, it is common ground in structuralist grammar, valency theory and functional grammar that it is the meaning of the ditransitive verb that determines the number and kind of semantic roles in the ditransitive complementation of the verb. I will therefore keep to the well-established understanding of transitivity as a primarily verb-related notion. Also, I agree with Meyer (1995) that a clear distinction should be made between the concepts of complement (i.e. the clause element ‘C’ in the clause patterns SVC and SVOC) and complementation (i.e. all clause elements required by the verb, which may be subject, object, adverbial and/or complement). More specifically, ditransitivity exemplifies a specific kind of verb complementation, but does not necessitate the use of a complement. Thus, I do not take over the way the term ‘complementation’ is used in CCEG either. Notwithstanding these terminological reservations, the CCEG resembles both the Comprehensive Grammar and functional grammars in defining ditransitive verbs in terms of the ‘participants’ (i.e. semantic roles) involved: Sometimes you want to talk about an event which involves someone in addition to the people or things that are the subject and object of the clause. This third participant is someone who benefits from the action or receives something as a result. They become the indirect object of the clause. ... Verbs which can take both a direct object and an indirect object are called ditransitive verbs. (Sinclair, 1990: 159) On this basis, the ditransitive complementation types [D1] and [D2], established in the Comprehensive Grammar, are described and illustrated with numerous authentic examples obtained from the COBUILD Corpus (Sinclair 1990: 159ff.): the basic form (e.g. Dad gave me a car); the indirect object realised as a prepositional phrase (introduced by to or for) and following the direct object (e.g. Ralph passed a message to Jack and He had saved an apple for the beggar respectively); ditransitive verbs which need not be complemented with an indirect object (e.g. He had left a note). Clausal realisations of objects are not covered. Of particular importance are the ‘usage notes’ for learners of English and the great number of word-lists containing verbs which occur significantly frequently in a specific complementation type. The main innovation of the functional and learner-oriented CCEG lies in the quantitative evidence for grammatical description: “This grammar attempts to make accurate statements about English, as seen in the huge Birmingham Collection of English Texts” (Sinclair 1990: v). No special emphasis is placed on the development and/or refinement of grammatical concepts, categories and terminology. The corpus serves as a new descriptive method and as the grammarian’s auxiliary equipment.47 47 This is not to be taken as a criticism. Sinclair has certainly been in the vanguard of the development of corpus linguistics as a linguistic discipline in its own right. Some of Sinclair’s numerous and innovative suggestions and concepts will be discussed in section 1.2.6. However, the CCEG remains a largely traditional grammar using the corpus as a
40
Chapter 1
Only at first glance does this appear to hold true for Biber et al.’s (1999) Longman Grammar of Spoken and Written English (LGSWE) as well: From CGEL [Comprehensive Grammar], we have also borrowed, with few exceptions, the grammatical framework of concepts and terminology which has provided the present book with its descriptive apparatus. While advances in corpus technology have allowed us to go beyond CGEL in important ways – particularly in the exemplification and quantitative investigation of grammar across different language varieties, spoken and written – CGEL’s attention to detail and comprehensive coverage is something to which this grammar does not attempt to aspire. In many ways, the two grammars complement rather than compete with each other. (Biber et al., 1999: viii) Notwithstanding the authors’ own modest assessment of the scope and coverage of their work, in my view the LGSWE is a completely new kind of grammar.48 Clearly, it has to be seen in the tradition of the Comprehensive Grammar from which many concepts and terms are taken over, but the underlying methodology is now entirely corpus-based and leads to new insights into actual language use. Specifically, the distinction of complementation types is to a large extent in accord with Quirk et al. (1985), but this descriptive framework is taken as a starting-point for an in-depth analysis of a 40-million-word corpus comprising written and spoken texts (of British English and American English). Not always do the authors systematically and exhaustively analyse the corpus in its entirety, though. Not infrequently, they confine themselves to subcorpora or small samples. Nevertheless, the quantitative findings are usually analysed from a functional point of view, from which inspiring and at times surprising source of authentic examples. This in itself is a great achievement since the CCEG is the first grammar to be solely based on actual data obtained from a large and representative corpus of spoken and written English. 48 It is somewhat surprising that reviewers of the LGSWE do not at all agree whether it really follows the descriptive framework provided by the Comprehensive Grammar and, consequently, to what extent the LGSWE is in fact innovative. For example, Oostdijk (2000: 16) insists on the conceptual similarities between the two grammars and, in her view, the negative result: “A point of weakness of the LGSWE, I think, is the extent to which this grammar capitalizes on the grammatical description in CGEL.” Krug (2002), too, stresses the similarities between the two grammars but draws a positive conclusion. On the other hand, Görlach (2000: 258) explicitly speaks of “a completely new work” and is very appreciative of the achievements of the LGSWE. Schneider (2001) holds an intermediate position, admiring the innovative character of the grammar but at the same time feeling uneasy about the lack of theoretical commitment and terminological clarity. Schneider (2001: 143) argues that these shortcomings are probably caused by the rapid compilation of the grammar, the need for speedy (and, thus, at times sloppy) work, being “indicative of an age dominated by the pressures of McDonaldization, globalization and commercialization – tastes good, digests easily, sells well.”
Ditransitive Verbs in Previous Research and in the Present Work
41
conclusions are drawn. The corpus data are grouped into four major genres, which allows for a comparative analysis of actual language use in ‘academic prose’ (ACAD), ‘conversation’ (CONV), ‘fiction writing’ (FICT) and ‘news writing’ (NEWS). Many aspects of the description of major clause patterns in the LGSWE are in accordance with the categories established in the Comprehensive Grammar. However, Biber et al. (1999: 141) place special emphasis on the term ‘valency’, “a term originally used in chemistry for the combinatorial potential of atoms”. Accordingly, they distinguish between ‘one-place verbs’, ‘two-place verbs’ and ‘three-place verbs’. Ditransitive three-place verbs fit into the clause pattern ‘subject – verb phrase – indirect object – direct object’ (Biber et al. 1999: 150), which corresponds to the basic form of ditransitive complementation [D1] in the Quirk-grammar. Biber et al. (1999) admit that most ditransitive verbs are open to other than the basic form of complementation as well: Notice that most ditransitive verbs also have ditransitive prepositional uses: compare the first example in this section [Well that tells you the voltage of the battery] with the last in 3.5.7 [He only told his name to an Italian painter named Carlino]. (Biber et al., 1999: 150) The important difference between the Comprehensive Grammar and the LGSWE lies in the fact that the latter grammar does not view verbs as ditransitive if they do not occur in the basic form of ditransitive complementation (e.g. three-place prepositional verbs such as accuse of). Thus, Biber et al. (1999) insist on the point that, irrespective of the variant forms of the clause pattern SVOO, a ditransitive verb must be attested in the basic form of the ditransitive clause pattern, i.e. SV[Oi:NP][Od:NP]: Ditransitive verbs occur with two object noun phrases – an indirect object and a direct object – in the pattern SVOiOd: [Fred Unsworth <S>] [gave ] [her ] [a huge vote of confidence ]. (Biber et al., 1999: 381) This formal criterion is of prime importance for the present study because it makes it possible to exhaustively identify all ditransitive verbs in a machinereadable corpus by way of automatic searches. As will be outlined in section 1.3.1, the somewhat fuzzy semantic component of ditransitivity, as illustrated in Figure 1-8, can thus be complemented with a formal aspect: in the present work, verbs are considered ditransitive only if they occur in the basic form of ditransitive complementation. It should be noted, though, that the definition of ditransitive verbs in the LGSWE also refers to all variations of ditransitive complementation once the aforementioned formal criterion is met by a given verb. Not only will I keep to this broad understanding of ditransitivity with the basic form at its basis, but I will also subsume only single-word lexical verbs into the category of
42
Chapter 1
ditransitive verbs, in this also following the LGSWE (see section 1.3.1). The exclusive focus on single-word verbs is, by the way, again in stark contrast to the Comprehensive Grammar, which also views many prepositional verbs as ditransitive (see section 1.2.1). With regard to clausal realisations of direct objects, the LGSWE largely capitalises on the descriptive apparatus of the Comprehensive Grammar. In a similar vein to Jackson (1990), Biber et al. (1999) also take into account contextdependent zero realisations of direct objects so that the following ditransitive complementation patterns are listed for the ditransitive verbs show and get:49 Ditransitive SVOiOd (NP): I want to show him the cover. (CONV) Why don't you go and get us both a pie. (FICT) Ditransitive SVOdOi (with indirect object in a prepositional phrase): Can I show this to Ian? (CONV) Noel brought out Michael's papers again and showed them to me. (FICT) Ditransitive SVOi + complement clause: Another glance showed me that it was carried on a stick by a man. (FICT †) Ditransitive with only an indirect object SVOi: Yeah I showed my mom. (CONV) I'll show you. (FICT) (Biber et al., 1999: 391) It should not go unmentioned that, similar to the Comprehensive Grammar, Biber et al. (1999: 152ff.) also briefly discuss possible ‘variations on clause patterns’, for example in passive constructions and object fronting. I do not intend at this stage to replicate the quantitative data on the use of ditransitive verbs to be found in the LGSWE. Whenever necessary, they will be taken into consideration in the course of the corpus analysis in chapter 3. From what has been said so far about Biber et al.’s (1999) general approach to ditransitivity, the conclusion can be drawn that they modify to a considerable extent the descriptive apparatus of the Comprehensive Grammar. Specifically, the definition of ditransitive verbs is much more restricted in scope since it includes a formal component not inherent in the Comprehensive Grammar. Thus, the definition of ditransitive verbs in the LGSWE is applicable to a computerised corpus analysis, because ditransitive verbs can now be automatically retrieved on grounds of the basic form of ditransitive complementation. It must be admitted that while the treatment of ditransitive complementation in the LGSWE differs significantly from the Comprehensive Grammar, the two grammars in their entirety are in fact similarly based on the descriptive and structuralist tradition. In general, then, the LGSWE represents in many regards a 49
The symbol ‘†’ indicates truncated examples.
Ditransitive Verbs in Previous Research and in the Present Work
43
corpus-based add-on to the Comprehensive Grammar. In principle, the foundations of the received structuralist framework are not called into question but are more or less modified and/or refined in the light of corpus data and modern computerised research possibilities. A much more radical approach to grammar has been suggested recently by proponents of corpus-driven lexicogrammar, to which the following section is devoted. Here, corpus data do not simply complement existing frameworks but motivate linguists to approach grammar from an entirely new perspective. 1.2.6 Corpus-driven lexicogrammar Sinclair (1991: 137) was to the fore in formulating the hypothesis that the received idea of a clear-cut distinction between lexis and grammar might prove untenable in the light of corpus evidence: “when we have thoroughly pursued the patterns of co-occurrence of linguistic choice there will be little or no need for a separate residual grammar or lexicon.” In fact, corpus research reveals that the positing of an integrated ‘lexicogrammar’ is much truer to linguistic reality because actual language use is based to a very large extent on lexicogrammatical routines, e.g. ‘colligations’ as co-selections of word-forms and grammatical categories (Sinclair 1996: 85).50 Starting off from Sinclair’s idea of a unified lexicogrammar, Francis et al. (1996, 1998) have produced two dictionaries of the patterns in which all nouns, adjectives and verbs listed in the Collins COBUILD English Dictionary (Sinclair 1995) occur significantly frequently. The two volumes (designed for learners of English in particular) also allow learners to look up all the words which frequently occur in a specific pattern. The overarching point here is that obviously lexis and grammar are interdependent to an amazing extent: specific words tend to occur in specific grammatical patterns, and specific patterns tend to select a semantically restricted range of words. This interrelation of lexis and grammar has led Hunston and Francis (2000) to develop a ‘pattern-grammar approach’ to English grammar. In this, the key notion of pattern is defined by Hunston and Francis (2000) as follows: The patterns of a word can be defined as all the words and structures which are regularly associated with the word and which contribute to its meaning. A pattern can be identified if a combination of words occurs relatively frequently, if it is dependent on a particular word choice, and if there is a clear meaning associated with it. (Hunston and Francis, 2000: 37) The pattern-grammar approach is a good example of an integrated lexicogrammatical perspective on language which starts off from corpus evidence obtained 50
Recently, Stefanowitsch and Gries (2003) have introduced the notion of ‘collostruction’, which captures the mutual attraction of a lexeme and a construction to each other.
44
Chapter 1
from the Bank of English Corpus (with then some 329 million words). This approach is of particular importance for the present study since ditransitive verbs occur in a restricted range of ditransitive patterns and, at the same time, ditransitive patterns are associated with ditransitive verbs as a semantically defined type. Thus, there is a semantic component inherent in ditransitive complementation – an assumption which is clearly reminiscent of functional grammar (see section 1.2.4). In other words, from the point of view of pattern grammar, the combination of a ditransitive verb and its ditransitive complementation ought to be viewed as a lexicogrammatical unit of meaning in its own right. In order to represent patterns, Hunston and Francis use a small inventory of iconic symbols which follow the notation of the Collins COBUILD English Dictionary (Sinclair 1995: xxx ff.). For instance, the pattern ‘V n n’ refers to a specific verb which tends to be followed by two noun phrases as in I wrote him a letter (Hunston and Francis 2000: 53).51 With regard to the traditional functional categories of clause elements in descriptive grammars, this pattern occurs in the clause patterns SVOO and SVOC: Firstly, the traditional ditransitive verbs have this pattern, as in I gave him some bread and He made me a sandwich. Secondly, verbs followed by an Object and an Object Complement have the same pattern, as in She called them all idiots. (Hunston and Francis, 2000: 87) If a ditransitive verb occurs in this very pattern (with ‘V n n’ referring to the basic form of ditransitive complementation), all the verbs can be subsumed into one of the following ‘meaning groups’(cf. Hunston and Francis, 2000: 88f.): (a) giving someone something, or refusing to do so (e.g. give, deny); (b) doing something for someone (e.g. bring, play); (c) talking, writing, or otherwise communicating something to someone (e.g. tell, fax); (d) giving someone a benefit or a disadvantage (e.g. save, charge); (e) verbs concerned with feeling and attitudes (e.g. envy, forgive). These meaning groups make it clear that, at an abstract level, ditransitive verbs share similar semantic characteristics.52 What is more, Hunston and Francis provide empirical evidence for the fact that all ditransitive verbs attested in a 329million-word corpus display this semantic commonality. Note also that the five meaning groups above correlate to a considerable extent with Dixon’s semantic types (see section 1.2.4): the meaning groups (a) and (b) refer to the semantic 51
Since the pattern of the verb is under discussion, the corresponding symbol ‘V’ (for ‘verb’) is capitalised. 52 Note in this context that some of the meaning groups are less prototypical than others. The verbs that are included in these meaning groups, e.g. envy, have a meaning which is remote from the meaning of the prototypical ditransitive verb give.
Ditransitive Verbs in Previous Research and in the Present Work
45
type of GIVING, (d) and (e) to AFFECT, and (c) to SPEAKING. It seems to be a matter of delicacy, then, whether three or five semantic types or meaning groups should be distinguished. What both classification schemes have in common is the underlying semantic assumption that the ditransitive complementation in the pattern ‘V n n’ has a meaning of its own, which could be described as an event in which a concrete, abstract or verbal transfer takes place. In this regard, corpus data on verbs associated with the pattern ‘V n n’ corroborate the usefulness of the largely intuition-based assumption in functional grammar that ditransitive verbs should also be defined semantically. However, one point of weakness in the pattern grammar approach for the purpose of the present study is its focus on the surface structure in terms of formal patterns (such as ‘V n n’). Hunston and Francis (2000) deliberately abstain from an in-depth ‘structural analysis’ in terms of traditional functional categories (such as object and complement): In most cases ... the structural analysis added nothing, and all that was important to say about a verb could be said in terms of its pattern and its meaning group, irrespective of the structural interpretation. Our conclusion was that structural analysis is a pointless exercise ... . (Hunston and Francis, 2000: 152) However, the abandonment of a structural analysis leads to the problem that the gap between the merely syntactic description of patterns and the semantic analysis of meaning groups cannot be bridged. Needless to say, ditransitive complementation is not exclusively bound to the pattern ‘V n n’. The focus on this pattern disguises the fact that, for example, the direct object may also precede the indirect object (which would then be realised as a prepositional phrase). This variation of ditransitive complementation, in which, for instance, the verb give frequently occurs, would correspond to the pattern ‘V n to n’. Since functional categories such as the notion of object are ignored, there is no systematic account of all the ditransitive patterns with which a given ditransitive verb is associated. It is a clear disadvantage of this approach that a functional analysis of the (say, context or genre-dependent) factors which may lead the language user to prefer one specific ditransitive pattern of give to another is thus not possible. The learner-oriented pattern grammar account thus fails to give learners of English any information about when to use a specific ditransitive pattern with a given ditransitive verb (cf. Smitterberg’s (2000) critical remarks). In the aforementioned pattern dictionaries of English, Francis et al. (1996, 1998) do use functional categories. However, functional categories are not introduced in order to systematise the semantic and syntactic correspondences between different patterns. Rather, they are exclusively used to distinguish between different meaning groups within a given pattern, e.g. between ‘Structure I: Verb with two Objects’ and ‘Structure II: Verb with Object and Object Complement’ within the pattern ‘V n n’ (Francis et al. 1996: 272ff.). This does not alter the fact that each and every pattern is dealt
46
Chapter 1
with in isolation. Only as a last resort are functional categories used whenever semantic distinctions within a given pattern turn out to be impossible otherwise. A problem arising from the abandonment of functional categories relates to the syntax-semantics interface. Specifically, the range of formal realisations with which one and the same semantic role is associated is not systematised. For example, by considering the two patterns ‘V n n’ and ‘V n that’ in isolation, the pattern grammar approach does not capture the semantic correspondences between the realisation of a direct object as a noun phrase and as a that-clause. But the meaning of the ditransitive verb (in terms of its meaning group) does not change simply because one of the semantic roles it requires is realised differently. Furthermore, passive realisations are not treated as patterns in their own right in the pattern grammar approach. With regard to ditransitive verbs, Francis et al. (1996: 272), for example, list the pattern ‘be V -ed n’ as a variant form of the canonical pattern ‘V n n’, which appears to be problematic since some verbs (e.g. assign) are explicitly labelled ‘usu[ally] passive’. Why should in such cases the pattern ‘be V -ed n’ be a derivative of the pattern ‘V n n’ if the verb only scarcely occurs in the latter pattern? In sum, the abandonment of a structural analysis results in some fundamental shortcomings which prevent the no doubt innovative pattern-grammar approach from achieving its full potential. In any unified approach to lexicogrammar and semantics, I think that linguistic categories which link different surface patterns to their shared underlying meaning should be taken into consideration. In the present study, this is done by drawing on the concept of semantic roles at the level of abstract propositional meaning and by adhering to functional categories such as the notion of object at the level of syntax. Thus, I suggest that the three levels of semantic roles, functional categories and lexicogrammatical patterns be taken as representing three different, but related fields of analysis: (1) semantic roles refer to the participants involved in the underlying argument structure, but are not (yet) bound to a specific linear order of the corresponding syntactic clause elements; (2) the notions of subject and object already imply to a considerable extent information about the surface elements and a basic linear order of elements, but are not (yet) bound to a specific formal realisation; (3) lexicogrammatical patterns as concrete formal realisations and final linearisations (of semantic roles and of functional categories, that is) refer to the eventual product of language performance, which is accessible to linguistic observation. The levels refer to the basic steps in language production, which is generally believed to proceed in an incremental top-down process from semantic roles to formal realisations (see section 1.2.8). By means of an empirically sound, corpus-linguistic methodology, the three levels can be approached in a bottom-up process which makes it possible, first, to describe correspondences between all patterns in which ditransitive verbs occur significantly frequently (and to identify the relevant ditransitive verbs in the first place); second, to analyse from a functional and discourse perspective why under specific contextual circumstances one ditransitive pattern is preferred to others; third, to develop a plausible model of the cognitive entrenchment of ditransitive verbs and their patterns. The three levels of analysis
Ditransitive Verbs in Previous Research and in the Present Work
47
and the relevant terminology with regard to the object of inquiry at hand are summarised in Table 1-1. Table 1-1:
Levels of analysis and descriptive categories in an integrated approach to ditransitivity
Levels of analysis
Descriptive categories
Elements of descriptive categories
cognition/ semantics
situation type, semantic roles
A transfers B to C (GIVING, AFFECT, SPEAKING) acting entity, transferred entity, affected entity
basic linearisation of semantic roles and selection of clause elements clause elements
subject (nominal/clausal), ditransitive verb, object (direct/indirect/ prepositional/clausal)
formal realisation of clause elements and final linearisation in patterns lexicogrammar
ditrans. verbs, give, send, bring, forgive, tell, write etc. ditrans. patterns V n n, V n to n, V n that-cl., V n to-inf. etc.
ditransitive complementation
functional categories
Let me emphasise in this context that the corpus as a source of linguistic data is of pivotal importance for the present study because it allows – within a given corpus – for an exhaustive and empirically solid analysis of ditransitive verbs and their patterns as well as for the description of contextual factors which play a role in the process of pattern selection. However, I think that the recurrent patterns obtained from corpus data (and on which Hunston and Francis place special emphasis) should be taken as the basis for further steps of a functional and semantic analysis of ditransitive verbs, as represented by the two upper levels in Table 1-1. The need for semantic and/or cognitive categories, which are in principle independent of formal lexicogrammar, is also emphasised by cognitive linguistics, to which I will turn in the following section. Focusing on Goldberg’s (1995) seminal work on construction grammar and Langacker’s (1999) cognitive grammar approach to ditransitivity, I will show that major assumptions of the two frameworks are perfectly compatible with structuralist, functional and corpuslinguistic concepts.
48
Chapter 1
1.2.7 Construction grammar and cognitive grammar Typological research has provided ample testimony to the fact that transitivity in general is a phenomenon that can be found in virtually all languages (cf. e.g. Hopper and Thompson, 1980; DeLancey, 1984 and 1987). Thus, it is reasonable to assume that transitivity represents a syntactic category which refers to a universal semantic prototype in terms of an all-pervading event type in the extralinguistic world. Accordingly, DeLancey (1987) puts forward the following hypothesis: I think it can be shown that the natural basis for the transitivity prototype is the universal human understanding of the physical fact that events have causes, i.e. that the basis for the transitivity prototype is a simple CAUSE EFFECT schema which owes its universality to its universal utility in dealing with the real world. (DeLancey, 1987: 60) Thus, an implicit link is established between the cognitive representation of the outside world in terms of event types (in the mind) and the universal syntactic category of transitivity (in language).53 What is more, it has also been suggested in particular that some kind of ditransitive structure, i.e. a grammatical construction for encoding a transfer of an entity from one entity to another entity, is also universal (cf. Newman, 1996 and 1998). In general, neither cognitive categories nor corresponding syntactic categories are believed to have distinct boundaries but are taken to be gradients of prototypicality (cf. e.g. Rosch 1978, B. Aarts 2004). I have already noted before that ditransitivity in particular can also be viewed both as a cognitive category in terms of the prototypical schema ‘A transfers B to C’ and as a syntactic category characterised by a specific (range of) verb complementation (cf. Figure 1-8 and Table 1-1). If we follow the assumption that ditransitivity is a category with central prototypical elements (e.g. ditransitive verbs which tend to occur in the basic form of ditransitive complementation), but which also comprises less prototypical elements (e.g. ditransitive verbs which tend to occur in other complementation patterns), then it is quite obvious that ditransitivity cannot be defined by focusing on a specific formal realisation (say, the pattern ‘V n n’) alone. Rather, traditional concepts and terminologies of descriptive, functional and corpus-linguistic grammars need to be complemented with a cognitive component. This is exactly what Chafe (1987), among other things, suggests: It can be viewed as part of a continuing effort to become clearer on what is happening in people’s minds when they manipulate “given 53
The notion of ‘event type’ is used here in a non-technical sense. I do not intend to discuss a separate event-structural level of linguistic analysis as suggested, for example, by Pustejovsky (1992).
Ditransitive Verbs in Previous Research and in the Present Work
49
and new information”, “topics and comments”, “subjects and predicates”, as well as “intonation units”, “clauses”, “sentences”, and “paragraphs”. I assume that the linguistic phenomena which have been given names like these are manifestations of basic cognitive processes, and that we can never understand them fully until we understand the psychological phenomena underlying them. (Chafe, 1987: 21) In this context, two cognitive-linguistic approaches are particularly noteworthy because they have found wide acceptance in general and because they can be fruitfully applied to the description and analysis of ditransitive verbs in particular: construction grammar and cognitive grammar. Construction grammar is concerned with the syntax and semantics of grammatical constructions. It is thus concerned with a level of analysis that has been neglected in so-called slot-and-filler models of language such as generative grammar. As Tomasello (1998: 431) points out, “[o]ne of the defining features of modern-day generative grammar is the absence of constructions.” Recently, linguistic research into the construction level experiences a remarkable renaissance in many non-generative frameworks, for example in corpus linguistics, which focuses on recurrent patterns in actual language use (see sections 1.2.5 and 1.2.6). There is also a revival of constructions in cognitive linguistics. Fillmore (1985, 1988) has certainly been in the vanguard of this development.54 With his ‘construction grammar’, he has developed an innovative descriptive apparatus which proves highly suitable for the analysis of the cognitive implications of syntactic structures, i.e. their correlations with underlying argument structures. On the one hand, construction grammar shares with generative linguistics an interest in speakers’ competence, i.e. their ability to use an infinite number of expressions. On the other hand (and this clearly distinguishes it from generative grammar), construction grammar neither assumes a division between lexicon and syntax nor between form and meaning, which is described by Goldberg (1995) as follows: Lexical constructions and syntactic constructions differ in internal complexity, and also in the extent to which phonological form is specified, but both lexical and syntactic constructions are essentially the same type of declaratively represented data structure: both pair form with meaning. (Goldberg, 1995: 7) 54 Again, it would be beyond the scope of the present work to provide an exhaustive introduction to the construction grammar approach. After discussing a few central tenets, I will focus on its applicability to the description of ditransitive verbs. For general introductions to construction grammar see, for example, Fillmore, Kay and O’Connor (1988); for its relevance to information structure Lambrecht (1994); for its origins Fillmore’s (1968) seminal article on “The case for case”.
50
Chapter 1
In general, the notion of construction refers to all form-meaning pairs – from morphemes and lexemes to phrasal structures (e.g. let alone) and to clause patterns (e.g. SVOO).55 Goldberg (1995: 9ff.) gives a detailed and considered account of the advantages of a constructional approach over alternative approaches such as lexicosemantic and generative-transformational rules. In the following, I will, however, restrict myself to her analysis of ditransitive verbs and ditransitive constructions, which exemplifies many of the advantages – and one major disadvantage – of the construction account in general. The ditransitive construction is regarded by Goldberg (1995) as an example of a basic sentence type in English imparting an equally basic event type (a ‘scene’) in human cognition with an appropriate linguistic structure. This is captured by the general ‘Scene Encoding Hypothesis’:56 Scene Encoding Hypothesis: Constructions which correspond to basic sentence types encode as their central senses event types that are basic to human experience. (Goldberg, 1995: 39) In applying this working premise to the ditransitive construction, the following, closely related senses, of which the first one is considered the ‘central sense’, from which all other senses are derived, is suggested (cf. Goldberg, 1995: 38): (a) Central sense: agent successfully causes recipient to receive patient: - verbs that inherently signify acts of giving (e.g. give, pass) - verbs of instantaneous causation of ballistic motion (e.g. throw, toss) - verbs of continuous causation in a deictically specified direction (e.g. bring, take) (b) Conditions of satisfaction imply that agent causes recipient to receive patient: - Verbs of giving associated satisfaction conditions (e.g. guarantee, promise) (c) Agent causes recipient not to receive patient: - Verbs of refusal (refuse, deny) (d) Agent acts to cause recipient to receive patient at some future point in time: - Verbs of future transfer (e.g. leave, bequeath) (e) Agent enables recipient to receive patient: - Verbs of permission (permit, allow) (f) Agent intends to cause recipient to receive patient: - Verbs involved in scenes of creation (e.g. bake, make) - Verbs of obtaining (e.g. get, grab) 55
As defined by Goldberg (1995: 4), “C is a CONSTRUCTION iffdef C is a form-meaning pair such that some aspect of Fi or some aspect of Si is not strictly predictable from C’s component parts or from other previously established constructions.” 56 As Goldberg (1995: 39) points out, this “idea that constructions designate scenes essential to human experience is reminiscent of Fillmore’s original motivation for the existence of a particular, fixed set of case roles” – a concept which is very much in accordance with the semantic roles suggested in functional grammar (see section 1.2.4).
Ditransitive Verbs in Previous Research and in the Present Work
51
To a certain extent, this list is in line with the semantic interpretation of the ditransitive clause pattern SVOO in functional grammar and in corpus linguistics (see sections 1.2.4, 1.2.5 and 1.2.6).57 Note in this context that the ditransitive construction is regarded as a prototypical category which also includes verbs that cannot be easily mapped onto – but are still taken to be related to – the semantic core of ditransitivity, e.g. deny. This prototypicality approach makes it possible to also link verbs such as envy, (be)grudge and spare, which according to Gropen et al. (1989: 241) could be subsumed under ‘future not having’, to the ditransitive transfer event.58 The important thing in Goldberg’s list of senses that can be ascribed to the ditransitive construction is the underlying interaction between verbs and constructions. Following the Scene Encoding Hypothesis, the ditransitive construction represents a more or less specific argument structure, that is an event type with a certain number and kind of ‘argument roles’: i.e. ‘agent’, ‘recipient’ and ‘patient’. In Goldberg’s (1995: 49) notation, the argument structure would be given as ‘CAUSE-RECEIVE ’. At the level of argument structure, then, the construction itself carries meaning, which corresponds to the meaning of the lexicogrammatical pattern ‘V n n’ as suggested by Hunston and Francis (2000). It is here that the interaction between verb and construction comes into play. The basic assumption is that only verbs can be used in this construction which allow for a ‘fusion’ between the ‘participant roles’ of the verb (e.g. ‘giver’, ‘receiver’ and ‘thing given’ for the verb give) and the argument roles of the construction. As Goldberg (1995) states, the fusion of participant roles and argument roles is subject to routinisation: If a verb is a member of a verb class that is conventionally associated with a construction, then the participant roles of the verb may be semantically fused with argument roles of the argument structure construction. (Goldberg, 1995: 50) Apart from classic ditransitive verbs, it is now possible, by drawing on the concept of fusion between verbal participant roles and constructional argument roles, to provide a powerful explanation for the fact that only specific novel verbs, which are not traditionally associated with the ditransitive construction, 57
Note that the classification of ditransitive verbs into the six categories is based on shared similarities with regard to the number and kind of the ‘participant roles’ they require. Goldberg (1995: 135) also offers a rough-and-ready and illuminating visualisation of such ‘similarity clusters’. In a wider setting, her convincing list of the senses associated with ditransitive verbs calls into question previous assumptions that a characterisation of all ditransitive verbs in terms of shared semantic roles is not possible (cf. e.g. Hudson 1992: 256). 58 Goldberg (1992: 67) also points out that “[f]orgive and envy historically had senses that were closely related to give”, which shows that “the construction can occasionally be frozen without continuing reference to the original semantics.”
52
Chapter 1
can be used in this very construction. One example which Goldberg (1995: 53) discusses is the integration of the verb mail with the ditransitive construction.59 Obviously, ditransitivity is both a lexical and a constructional (and both a linguistic and a cognitive) phenomenon: the basic idea is that only if the lexical item and the construction at hand are semantically compatible is the lexicogrammatical product acceptable. The process of fusion is principle-guided: Goldberg (1995: 50f., 61ff.) suggests ‘The Semantic Coherence Principle’ (referring to the qualitative compatibility of participant roles and argument roles), ‘The Correspondence Principle’ (referring to the quantitative compatibility of participant roles and argument roles) and the ‘Causal Relation Hypothesis’ (referring to the need for a causal relationship both in verbal and in constructional meaning). Major tenets of construction grammar can be easily reconciled with functionalist (and, indeed, traditional and structuralist) approaches to syntax and semantics: verbal ‘participant roles’ and constructional ‘argument roles’ correspond, grosso modo, to the functional categories and related semantic roles respectively that are required by the verb in its complementation. The ditransitive construction itself corresponds to – but is not at all identical with – the notion of lexicogrammatical pattern which has been suggested by proponents of corpusdriven grammar. The main and no doubt invaluable innovation of construction grammar lies in the development of a cognitive framework of argument structure which can be readily mapped onto previous research into the syntax and semantics of ditransitivity. Thus, Goldberg’s construction grammar approach to the ditransitive construction corroborates, from a more cognitive and argumentstructural point of view, the usefulness of a functional and integrated analysis of ditransitive verbs and ditransitive complementation. However, there is one major theoretical and a related descriptive problem in Goldberg’s approach which makes it difficult to base the present study on construction-grammar concepts. The theoretical problem concerns the treatment of syntactic variations of ditransitive complementations, most notably with regard to the relationship between the basic form of ditransitive complementation and the use of a prepositional paraphrase, for example in John gave Mary an apple vs. John gave an apple to Mary. In fact, Goldberg (1995) rejects the assumption that the correspondence between the two forms should be dealt with at the level of syntax: The question that arises, on the account presented here, is not whether verbs are allowed to undergo a lexical or syntactic rule that alters their semantic structure or subcategorization frame, as it is typically taken 59 It should be noted in passing that her cluster-based approach to the productive extension of the ditransitive construction to novel verbs (cf. Goldberg, 1995: 133ff.) differs considerably from Gropen et al.’s (1989) lexicon-oriented concept of the learnability of ditransitive syntax. However, issues of language acquisition are not immediately relevant to the present study. For details of the acquisition of the ditransitive construction see, for example, Tomasello (2000), Campbell and Tomasello (2001).
Ditransitive Verbs in Previous Research and in the Present Work
53
to be. Rather, the question becomes: How are the semantics of the independent constructions related such that the classes of verbs associated with one overlap with the classes of verbs associated with another? (Goldberg, 1995: 89) From a corpus-linguistic perspective, I feel that the question should rather be turned upside down: the different constructions at hand are not a priori independent but syntactically related to each other because many (if not all) ditransitive verbs can be used in different ditransitive complementation patterns, e.g. with both objects realised as noun phrases (and the indirect object preceding the direct object), with the indirect object realised as a prepositional phrase (and following the direct object), with either object becoming the subject of a passive construction. Irrespective of the concrete realisation of the ditransitive complementation, the meaning of the ditransitive verb in terms of the semantic roles it requires remains unchanged (although contextually reconstructible or irrelevant roles need not be expressed). In my view, the point of departure for an analysis of ditransitivity should be the ditransitive verb in all its ditransitive patterns. This is represented in Table 1-1 (p. 47) at the bottom level of analysis (i.e. the level of lexicogrammatical patterns), from which, in a bottom-up process, syntactic, semantic and, eventually, cognitive conclusions should be drawn. In construction grammar, on the other hand, the cognitive argument structure is taken as the descriptive starting-point. This, for example, leads Goldberg (1995: 91) to start off by discussing what she believes to be different constructions, e.g. the ‘ditransitive construction’ (e.g. in John gave Mary an apple with the agent as subject, the recipient as object, and the patient as second object) and the ‘transfer-causedmotion construction’ (e.g. in John gave an apple to Mary with the agent as subject, the recipient as oblique function and the patient as object).60 With the help of semantic extensions at the level of constructions, she then shows that the “caused-motion construction is (S)emantically synonymous with the ditransitive construction” (Goldberg, 1995: 91). In light of the fact that many other nonditransitive verbs are used in the clause pattern SVOA, I see no reason for postulating a merely cognitive link between the two ‘constructions’ – corresponding to the clause patterns SVOO and SVOA – because it clearly depends on the verb as such whether the two different complementations of a given verb are semantically synonymous (and syntactically possible in the first place). As a matter of fact, Goldberg’s account is plausible if one accepts the basic assumption that ditransitivity is a primarily cognitive (or constructional) phenomenon of which ditransitive syntax is symptomatic, as it were. However, it is difficult to capitalise on this strictly cognitive definition in a corpus-based study, whose first step is bound to the search for formal linguistic structures. An exhaustive corpus 60
In Stefanowitsch’s (2003: 418) terminology, this would mean that the two syntactic forms at hand are viewed not as ‘discourse-functional’ variants, but as representations of two distinct ‘information-structure constructions’.
54
Chapter 1
analysis on the basis of construction grammar seems to be possible only if the corpus is heavily semantically annotated. For this reason, I will adhere to the alternative (and equally plausible) view that at the level of cognition, there not only exist abstract constructions with implicit argument structures, but also – and more specifically – ditransitive verbs and their immediate associations with ditransitive patterns. This alternative view is in accordance with ‘usage-based models’ which are also intended to represent the cognitive basis of language. Closely related to the theoretical problem is the descriptive problem. The focus on argument structures and their cognitive interrelations is promising, but this focus leads Goldberg to ignore virtually all factors and principles which are at work in real speakers’ preference of one constructional variant over others under specific contextual circumstances. At the level of argument structure alone, functional and discourse-related interpretations of naturally occurring data are not possible. As Chafe (2000: 11) states, “[s]uch interpretations must in the end involve prosody, word order, givenness and newness, and discourse context, none of which Goldberg mentions”. As things stand, it turns out to be difficult to apply the construction grammar approach to a corpus-based analysis of authentic data. Nevertheless, it is rather amazing that there are quite a few cognitive linguists from mostly non-generative fields of research who emphasise the need for models of language cognition that take into account frequencies and distributions in actual texts.61 Such models are called ‘usage-based models’ by Kemmer and Barlow (2000): This idea of the fundamental importance of frequency … sharply distinguishes usage-based models from other approaches in which frequency is an insignificant artifact, unconnected with speakers’ linguistic knowledge. (Kemmer and Barlow, 2000: x) The term ‘usage-based model’ was first introduced by Langacker (1987) in order to describe his ‘cognitive grammar’ approach, which he characterises as follows: Substantial importance is given to the actual use of the linguistic system and a speaker’s knowledge of this use; the grammar is held responsible for a speaker’s knowledge of the full range of linguistic conventions, regardless of whether these conventions can be subsumed under more general statements. [It is a] nonreductive approach to linguistic structure that employs fully articulated schematic networks and emphasizes the importance of low-level schemas. (Langacker, 1987: 494)
61
A notable exception is Wasow (2002), who argues that generative theories need to be confirmed empirically by taking into account actual data and probabilities.
Ditransitive Verbs in Previous Research and in the Present Work
55
Cognitive grammar is based on a central counter-argument to generative grammar. Langacker (2000: 2) points out that, in his view, generative approaches suffer from the so-called ‘rule/list fallacy’: “Traditionally, in generative accounts, the instantiating expressions would be excluded from the grammar on grounds of economy.” That is to say that especially in the Minimalist Program, which is guided by economy conditions (see section 1.2.2), the underpinning assumption is that a maximally economical model must be based on an autonomous set of syntactic rules – i.e. a ‘computational’ system – on the one hand and a list of lexical entries on the other: “The I-language consists of a computational procedure and a lexicon” (Chomsky, 2000: 120). In fact, the attempt to separate a set of autonomous syntactic rules from a list of lexical entries has always, mutatis mutandis, been a cornerstone of generative theories. Langacker (2000) takes issue with this central generative argument and suggests that the two aspects be combined because this would increase the scope and plausibility of the model of grammar: There is a viable alternative: to include in the grammar both the rules and instantiating expressions. This option allows any valid generalizations to be captured (by means of rules), and while the descriptions it affords may not be maximally economical, they have to be preferred on grounds of psychological accuracy to the extent that specific expressions do in fact become established as well-rehearsed units. Such units are cognitive entities in their own right whose existence is not reducible to that of the general patterns they instantiate. (Langacker, 2000: 2) This suggestion, of course, cannot be reconciled with the clear-cut dichotomy of syntagmatic rules and paradigmatic lists in generative approaches. Interestingly enough, Langacker’s view is very much in line with corpus-driven lexicogrammar, which also puts special emphasis on lexicogrammatical units of meaning (see section 1.2.6): “My point, of course, is that lexicon and grammar grade into one another so that any specific line of demarcation would be arbitrary” (Langacker, 1999: 122). Not only does this statement resemble recent corpus-linguistic concepts, but also traditionally established positions in British contextualism such as the Hallidayan notion of ‘lexicogrammar’: “I have always seen lexicogrammar as a unified phenomenon, a single level of ‘wording’, of which lexis is the ‘most delicate’ resolution” (Halliday, 1991: 31f.). In principle, Langacker offers a cognitive framework which is intended to systematise such lexicogrammatical patterns. Of the wealth of relevant concepts and terminologies, the notion of ‘constructional schema’ is of particular interest for the present work because it captures the cognitive link between a ditransitive verb and its complementation: ... there is no evident reason why a constructional schema that incorporates a specific element, e.g. [[send][NP][NP]], should not also be a lexical item. That in turn is only one step away from according
56
Chapter 1 lexical status to assemblies like [[V][NP][NP]], all of whose components are schematic. (Langacker, 2000: 33)
Thus, ditransitive verbs are said to be cognitively associated with their complementation patterns, and patterns in return are at the same time cognitively associated with the verbs they select. This results in a lexical and constructional network that is intended to visualise the cognitive entrenchment of verbs, patterns and their interrelations. In Figure 1-9 below, one of Langacker’s (1999: 123) examples is given.
Figure 1-9:
Lexical and constructional networks in cognitive grammar (Langacker, 1999: 123)
In Figure 1-9, the constructional network (i.e. [[V][NP][NP]]) is in the left-hand circle, the lexical network (of send) in the right-hand circle and the resulting pattern (i.e. [[send][NP][NP]]) is located at the intersection of the two circles.62 The conceptual similarities between Langacker’s cognitive grammar and corpusdriven lexicogrammar are obvious, even though the objects of inquiry, namely 62 In Figure 1-9, the boxes with bold lines in the left-hand ellipse refer to constructional schemas, i.e. schemas with lexical items that are habitually associated with the ditransitive construction [[V][NP][NP]]: they include TRANSFER verbs in general and the ditransitive verb give in particular. On the other hand, the boxes with thin lines in the right-hand ellipse exemplify lexical schemas that are conventionally associated with the ditransitive verb send.
Ditransitive Verbs in Previous Research and in the Present Work
57
language cognition and language use respectively, are fundamentally different. For one, ditransitive patterns are entrenched in constructional networks (e.g. [[V][NP][NP]]). Also, ditransitive verbs are cognitively embedded in lexical networks which link the verbs to the patterns with which they are habitually associated (e.g. [[send2][NP][to NP]]. On the surface of intersection, correspondences between constructional networks and lexical networks are visualised (e.g. [[send][NP][NP]]).63 It should not go unmentioned that the pattern in the constructional network in Figure 1-9, i.e. [[V][NP][NP]], which corresponds to the pattern ‘V n n’ and the basic form of ditransitive complementation [D1], is associated with a semantic prototype. This prototype is described by Langacker (2000: 33) as ‘TRANSFER’ and can be specified by many different verbs: “Subschemas specify the occurrence of particular verbs in this pattern, give and send of course being common and well-entrenched”.64 In sum, the semantic concept of ditransitivity suggested in functional and corpus-linguistic grammar is in accord with Langacker’s characterisation of the cognitive aspects of ditransitivity. Furthermore, there are many useful points of reference in cognitive grammar for a corpus-based methodology. In particular, the notion of constructional schema correlates with the frequency-based concept of a conventionalised ‘linguistic schema’ as put forward by Barlow and Kemmer (1994) and Barlow (1996) and with the hypothesis of ‘lexicalized sentence stems’ as proposed in Pawley and Syder’s (1983) seminal article on the relevance of linguistic routine to native-like language use. In two major regards, cognitive grammar may profit from corpus-based methodology, which I would like to explain by drawing on Figure 1-9. First, a usage-based model should be based on actual language use. Thus, it will be shown in chapter 4, which will centre around cognitive aspects of ditransitivity, how corpus data may be included in constructional and lexical networks. For example, in Figure 1-9 nothing is said about the frequencies in which the verb send occurs in each of its patterns. Quantitative data obtained from large and representative corpora – and their careful analysis – provide useful clues in this context. It should be the overall aim of a usage-based model to offer a model which is derived from – and thus mirrors – authentic usage: high-frequency phenomena ought to be ascribed a different status in such a model than linguistic structures which occur less frequently or not at all. Second, the present work aims at a model which represents the cognitive entrenchment of ditransitive verbs in all its complexity and variety. For example, in Figure 1-9 ditransitive patterns with 63
Langacker (2000: 35) explicitly emphasises the theoretical importance of this intersection between lexical networks and constructional networks: “What, then, is the status of [[send][NP][NP]]? Does it belong to the ditransitive construction or to the lexical item send? The answer, of course, is that the question is wrong: it is simultaneously part of both.” Note also that Langacker's intersection is reminiscent of Goldberg’s (1995) concept of ‘fusion’ between verb and construction. 64 Again, Langacker’s notion of TRANSFER resembles Goldberg’s (1992: 70) ‘central sense’ of ditransitive syntax: “In the central sense, the argument structure is associated with a scene of transfer between a volitional agent and a willing recipient.”
58
Chapter 1
clausal realisations are not covered. In order to capture all possible formal realisations of the semantic roles involved in ditransitive complementation it is therefore useful to keep to the traditional concepts of functional categories, i.e. clause elements such as the direct/indirect object. This intermediate level of analysis indicated in Table 1-1 (p. 47), intended to mediate between cognitive semantic roles and lexicogrammatical patterns, is not explicitly mentioned by Langacker, but it can be easily integrated into his descriptive framework for he exclusively rejects only the following: “There is consequently no reason in this framework to hypothesize a level of representation (distinct from surface structure) at which constituency is fixed and invariant” (Langacker, 1999: 151). Clause elements may be bound to a specific linear order of elements to a considerable extent (except for cases such as object fronting), and may thus be considered more or less ‘fixed’, but they are not bound to a specific formal realisation of the underlying semantic roles, and are thus clearly not invariant.65 In conclusion, it makes sense to combine the corpus-based description of actual language use with Langacker’s cognitive grammar approach. In a wider setting, the present work is thus intended to bridge the gap between the close inspection of performance data and the modelling of speakers’ ‘competence’ – though quite differently from the generative notion of competence (see section 1.3.3). This general and theoretical objective will be achieved by analysing ditransitive verbs in usage and deducing from this analysis aspects of a cognitive grammar of ditransitive verbs. 1.2.8 Cognitive sciences Neurobiological and psycholinguistic research has been focusing on cognitive aspects of language for several decades now. Particularly noteworthy progress has been made in the fields of natural language processing and cognitive psychology. Some linguistically relevant results obtained from these lines of research will be discussed in this section. However, it is necessary to point out beforehand that, despite the rapid development of the cognitively-oriented disciplines, any cognitive model based on psycholinguistic experiments is still to a large extent hypothetical and speculative in nature. For example, the neurological essence of a linguistic rule or of an entry in the mental lexicon still defies direct empirical analysis. By means of conducting plausible experiments, it may be possible to get some indirect clues to the cognitive basis of linguistic phenomena, but hard evidence (in terms of neurobiological facts) for any concrete cognitive model still lies ahead. Accordingly, MacWhinney (2000) critically comments on the basic Chomskyan assumption that generative analyses are able to unveil the psychological reality of ‘internalised language’:
65
A reliable answer to the obvious question whether or not such functional categories exist in the psychological reality of language has not yet been given. There are contradictory findings (cf. e.g. Frazier and Fodor, 1978 and Tanenhaus et al., 1995).
Ditransitive Verbs in Previous Research and in the Present Work
59
Two basic observational problems faced by all of these analyses are the fact that no developmental psychologist ever observed a child learning a rule and that no neuroscientist ever traced the neural substrate of either a rule or a symbol. ... Of course, one could argue that the fact that no one has ever seen the top quark should not prevent us from constructing theories that rely on the reality of this subatomic particle. But analogies of this type are misleading. In fact, carefully controlled experiments with huge collectors of heavy water sunk deep in caves have provided solid tangible evidence for the reality of even this most elusive of physical entities. No such solid evidence has ever been provided for either linguistic rules or linguistic symbols. (MacWhinney, 2000: 122f.) Since solid biological evidence for a cognitive model of language is not (yet?) available, the present work suggests a radically different perspective on language cognition. In focusing on ditransitive verbs, my general stance on the issue is that empirically sound evidence for a usage-based model of language cognition can be found in large and representative corpora comprising spoken and written texts from actual usage. This theoretical aspect of the present study will be discussed in more detail in section 1.3.3 and in chapter 2. It will be shown that major tenets of such a usage-based model derived from – and based on – corpus data are in line with important concepts developed in the cognitive sciences. Any attempt to model language in the mind has to face the central problem of how to relate abstract cognition to concrete language data. This issue boils down to what Bock et al. (1992: 150) call the question of the “workings of the thought-to-language mapping process”. Two fundamentally different positions have been held in this regard: ‘mediated mapping’ versus ‘direct mapping’. Mediated mapping refers to the basic assumption that cognitive categories may be immediately related to a specific surface structure, but that alternative surface structures (e.g. the passive) are derived from a particular basic surface structure which is believed to be entrenched as the canonical form (e.g. the corresponding active sentence). It is obvious that this hypothesis underpins the generative line of argumentation: In this scheme, the deep-structure-object argument may become the surface-structure-subject of a passive through a movement operation. Because empty categories are the surface-structure vestiges of movedfrom deep-structure positions, this leaves a postverbal noun-phrase trace in the position from which the object was moved (The book was put [t] on the shelf). (Bock et al., 1992: 152) By drawing on psycholinguistic experiments, Bock et al. (1992: 159ff.) seek to verify or falsify the viability of the mediated-mapping hypothesis. In this, they focus on the linguistic realisation of the subject position. Roughly 200 informants
60
Chapter 1
took part in a recognition memory and verbal repetition test, in which a set of priming sentences (i.e. syntactic templates in active and passive voice) had to be matched with audiovisual input material. In the last resort, the experiment was designed to find evidence for or against the generative concept of active-passive transformations at the level of syntactic surface structures. To this end, the presented events involved both animate and inanimate agents and patients. The results, in general, call into question any clear-cut divison between deep-structural argument roles and surface-structural syntactic categories: The data show, first, that there was a propensity to bind semantically arguments to the same syntactic relations across successive, unrelated sentences. ... Second, the results show that there was a propensity to build similar structural configurations across successive, unrelated sentences. (Bock et al., 1992: 162) That is to say that in successive events that were to be verbalised, informants tended to consistently link the agent of an action to the subject position and to use either active or passive forms throughout – irrespective both of the animacy of the agent and of the available set of syntactic templates. As a matter of fact, these results have to be taken with a measure of caution since they are based on entirely artificial data elicited in experimental situations. Notwithstanding this reservation, they might still be regarded as an illuminating hint at the nature of underlying cognitive mechanisms in authentic and real-time speech production. It seems as though semantic roles are immediately linked to a specific syntactic slot, e.g. the agent as the logical subject to the position either of the grammatical subject or of the by-agent. This, however, would run counter to the mediated-mapping hypothesis. Rather, the results support the alternative view of the thought-to-language mapping process, which Bock et al. (1992) call ‘direct mapping’: Surface structures are generated ‘directly,’ without the mediation of deep structures, using a variety of nontransformational formal devices. One such device allows the subject and direct object and other syntactic relations to be stated with respect to different forms of verbs, generalizing the construct of lexical subcategorization to capture the fact that verb forms differ systematically in their placement of various arguments. (Bock et al., 1992: 152) Thus, the direct-mapping hypothesis can be applied to ditransitivity under the assumption that the different ditransitive patterns of a given verb (‘verb forms’), are independently represented in cognition. Bock et al. (1992: 152) refer to the following two verb forms of the verb give, which are equally and independently accessible to speech production from the outset:
Ditransitive Verbs in Previous Research and in the Present Work (9) (10)
61
giveprepositional dative [–subject give –direct object –oblique object] givedouble-object dative [–subject give –oblique object –direct object]
According to the direct-mapping hypothesis, the verb forms in (9) and (10) are not subject to surface-structure transformations but are stored separately: the verb form in (9) is associated with the prepositional dative and would be activated for sentences such as John gave the flower to Mary; the verb form in (10) is linked to the double-object dative and would be activated for sentences such as John gave Mary a flower. One side-effect of the direct-mapping hypothesis (with different verb forms associated with a given ditransitive verb) is that the generative-transformational issue of dative shift does not arise. In speech production, one of the verb forms of give is selected directly, e.g. (9) or (10), which automatically leads to different surface structures such as John gave a book to Mary versus John gave Mary a book. In a wider setting, a specific verb is not conceived of as one abstract lexeme that enters the mental lexicon, but as a bundle of individual verb forms, all of which are associated with different lexicogrammatical patterns or constructions and all of which are cognitive entities in their own right (cf. e.g. Lapointe 1985, Lapointe and Dell 1989). This view is, by the way, also corroborated by studies in error analysis (cf. e.g. Stemberger 1985).66 Langacker’s (1987, 1999, 2000) cognitive grammar also capitalises on a direct mapping between cognition and language: clearly, the different verb forms of a verb correspond to what can be found in Langacker’s constructional and lexical networks (see section 1.2.7). Despite the caveat that laboratory data should always be taken with a pinch of salt, Bock et al.’s (1992) results seem to vindicate the close interrelation of syntax and semantics and make it clear that a major generative claim is increasingly open to doubt, namely what Newmeyer (1998), among others, calls the ‘autonomy of syntax’:67 ... a system whose primitive terms are nonsemantic and nondiscoursederived syntactic elements and whose principles of combination make no reference to system-external factors. (Newmeyer, 1998: 23) The autonomy of syntax is incompatible with network-based models which integrate syntax and semantics. In the field of lexical semantics, such networks are, for example, suggested by Spitzer (1993), Maier and Spitzer (1999). They 66
Note also that from corpus-linguistic studies similar conclusions have been drawn. For instance, Sinclair (1991: 8) states that “each distinct [word-]form is potentially a unique lexical unit”. Drawing on this idea, Esser (2000a: 97) defines a ‘lexical linguistic sign’ as “the union of a single sense and a set of medium-independent, abstract grammatical wordforms. The set may include all or only a subset of the possible morphological forms.” 67 Other experimental studies also show that rather than the verb as a lexical item it is the individual patterns of one and the same verb that should be taken as cognitively entrenched entities; such a pattern-based model would also make it possible to include frequency information on the individual patterns of a given verb (cf. e.g. McElree, 1993).
62
Chapter 1
conclude from experiments with patients suffering from schizophrenia that lexical memory is made up of lexical units which are associated with each other both systematically and randomly. The model of such an ‘associational (semantic) memory’ is guided by two principles: (1) systematic semantic similarities between lexical units; (2) individual and random associations based on the entirety of previous linguistic experience. On the one hand, these studies point to the need for models of cognition which are network-like (and which are not based on a rigid division between rules and lists).68 On the other hand, they reveal that semantically similar verbs such as ditransitive verbs are probably not only associated with each other in the mind due to semantic similarities, but that these associations are stabilised through linguistic experience, i.e. by the fact that such verbs are used in similar constructions: actual usage is, thus, taken to play a major role in cognition. In a wider context, many frameworks developed in the cognitive sciences may help shed new light on the processes involved in the linearisation of cognitive entities and linguistic elements in language production. The problem of linearisation refers to the question of when and how elements are given a specific linear order during the production process. This particular problem is closely connected with the more general question when and how elements (cognitive, semantic and linguistic) are selected and structured on their way from the preverbal cognitive representation of a given state of affairs to the linguistic product. In a similar vein to Table 1-1 (p. 47), Habel and Tappe (1999: 119) suggest three different levels, which are intended to represent the three relevant steps to be taken in proceeding from the initial ‘multimodal representation of states of affairs’ over the ‘linear structuring’ of selected elements to the final ‘grammatical structure’ (see Table 1-2 below). Habel and Tappe, too, draw their conclusions from experimental data. Specifically, they analyse how informants verbalise drawing events which are presented in real time on a computer-screen. It is surprising to see that very often subjects do not adhere to the chronological order of events when syntactically arranging the corresponding linguistic items. This finding is clearly at odds with Levelt’s (1989: 138) ‘principle of natural arrangement’: “For event structures, the natural order is the chronological order of events.” On the contrary, Habel and Tappe’s results show that there is no such absolute isomorphy between event structure and linguistic structure. Thus, it makes sense to suggest two levels of linearisation, as indicated in Table 1-2: at the level of cognition, entities to be verbalised are selected and arranged in an event structure which is usually in line with the chronological order (‘propositional representation’); at the level of verbalisation, the corresponding syntactic categories are formally realised and arranged in a syntactic order which may or may not be in line with the chronological order of events (‘natural language discourse’). Note that although the term ‘linear structure’ is used for the propos68 The wealth of network-based models in applied fields of research, most notably perhaps in clinical linguistics and natural language processing, is indicative of their increasing relevance and successful application to practical studies in linguistics (cf. e.g. Schade and Eikmeyer, 1998; Eikmeyer et al., 1999; Jescheniak, 1999).
Ditransitive Verbs in Previous Research and in the Present Work
63
itional representation in Table 1-2, linearisation is also part of the formal realisation at the level of ‘grammatical structure’. Table 1-2:
From preverbal cognitive representation to natural language discourse (Habel and Tappe, 1999: 119)
processes of production and comprehension
conceptualization, conceptual analysis
levels of representation
structural properties of the representations multimodal representation structural similarity of states of affairs to states of affairs in the external world ↓ ↑ propositional representation of the textual content
phonological, syntactic and semantic en/decoding
linear structure
↓ ↑ natural language discourse
grammatical structure
As shown in Table 1-1 (p. 47), it appears to be necessary and plausible to distinguish between the linear order of semantic roles at the level of cognition and the formal realisation (including aspects of linearisation) at the level of lexicogrammar. In the light of the direct-mapping hypothesis, it has to be borne in mind that a specific linear order of semantic roles already considerably constrains the range of possible linear orders at the level of surface structure. The studies discussed in this section provide ample testimony of the usefulness of Langacker’s non-generative, network-like and usage-based model of language cognition, for example in regard to his unified approach to lexis and grammar, the implicit interrelation of syntax and semantics and the consideration of different linear orders of elements at the level of cognition. 1.3
Ditransitivity in the present work
In the following, it is my intention to bring together those aspects of the frameworks reviewed in section 1.2 that are relevant to the present study and that can be conceptually integrated with each other. This will lead to a working definition of ditransitive verbs that unites lexicogrammatical and functional as well as cognitive aspects of ditransitivity (see section 1.3.1). I will then describe in more detail how the usage of ditransitive verbs will be analysed in the present study and to what extent corpus-based methodology will be utilised (see section 1.3.2). Finally, it will be shown that also any model of language cognition can profit considerably from the results of an in-depth corpus analysis (see section 1.3.3).
64
Chapter 1
1.3.1 A pluralist theory of ditransitivity As summarised in Table 1-1 (p. 47), I regard three levels of analysis as relevant to the linguistic description of ditransitive verbs in actual usage: (1) cognition/ semantics (in terms of the ditransitive situation type with its semantic roles); (2) functional categories (in terms of clause elements); (3) lexicogrammar (of ditransitive verbs and patterns). The three levels cannot be separated from each other altogether and do overlap. Table 1-3 gives an overview of the three levels of analysis, the corresponding phenomena and the relevant categories. Table 1-3: level of analysis cognition/ semantics functional categories lexicogrammar
Aspects of ditransitivity at three levels of analysis inherent ditransitive phenomenon ⇒ ditransitive verbs are semantically similar in that they all require three semantic roles in the situation type TRANSFER ⇒ all ditransitive patterns represent the clause pattern SVOO with the obligatory clause elements subject, verb, indirect object and direct object ⇒ ditransitive verbs are associated with ditransitive complementation patterns
relevant categories • situation type • semantic role • • • • • •
subject (S) ditransitive verb (V) indirect object (Oi) direct object (Od) ditransitive verb ditransitive pattern
The levels and concepts listed in Table 1-3 represent the pluralist approach of the present work. It is an integrated framework which takes into account various relevant categories of ditransitivity, irrespective of the origins of the related concepts. Thus, the cognitive/semantic, the functional-grammatical and the lexicogrammatical categories of ditransitivity will all be considered (but, for obvious reasons, taken up in turn). In my view, neither a merely lexical or syntactic approach nor a strictly cognitive/semantic or functional-grammatical (or even pragmatic) approach is able to come to terms with the phenomenon of ditransitivity in all its versatility and complexity. My overarching claim is that linguistic phenomena such as ditransitivity are, of course, approachable from different perspectives: it is up to the linguist whether to put special emphasis, say, on formal or on functional aspects, on the lexicogrammatical patterns of ditransitive verbs or on the inherent semantics of ditransitive verbs. In order to achieve a fairly complete picture of the phenomenon at hand, it is, however, necessary not to lose sight of all the aspects involved. It has to be borne in mind that linguistic models are by definition – and in a similar vein to models in natural sciences – reasonable simplifications of reality. This, however, should not lead to oversimplifications as, for example, in strictly formalist (let alone, exclusively syntactic) models which manage without the functional side of ditransitivity altogether. On the other hand, a cognitive or semantic approach to ditransitivity
Ditransitive Verbs in Previous Research and in the Present Work
65
must always be based on the careful analysis of linguistic forms in usage. The pluralist approach to ditransitivity in the present work is aimed at combining useful concepts from various frameworks so that both formal properties of ditransitivity (in terms of ditransitive verbs and patterns) and functional aspects of ditransitivity (in terms of semantic and pragmatic factors) can be accounted for. At the basis of this pluralist approach to ditransitivity lies an equally multi-layered definition of ditransitive verbs:69 (11)
A working definition of ditransitive verbs A ditransitive verb (DV) is a trivalent verb that requires a subject (S), a direct object (Od) and an indirect object (Oi) for a complete syntactic complementation. It is necessary for all clause elements to be realisable as noun phrases (NPs): this realisation (S:NP – DV – Oi:NP – Od:NP) is called the basic form of ditransitive complementation. If a verb is attested in the basic form of ditransitive complementation in actual language use, it is also considered a ditransitive verb in all other forms of complementation. All ditransitive verbs and ditransitive complementations are associated with an underlying proposition that represents the situation type of TRANSFER with three semantic roles involved: the ditransitive verb denotes an action in which the acting entity transfers a transferred entity to the affected entity.
Apart from the integration of different perspectives on ditransitivity (as given in the definition of ditransitive verbs above), the present work has two theoretical innovations to offer. First, the description of ditransitive verbs in actual language use will be entirely corpus-based. I will argue in chapter 2 that the linguistic corpus not only represents a new method of obtaining data but that corpus-based methodology is based on specific theoretical foundations (which are often ignored, however, even by corpus linguists) so that it makes sense to hold corpus linguistics to be a linguistic discipline and not just a method (see section 2.3). Second, the corpus-based quantitative and qualitative analysis of ditransitive verbs in chapter 3 will be taken as a basis in chapter 4 for the development of a usage-based model of ditransitive verbs in cognition. In this, the theoretical apparatus provided by cognitive grammar will be complemented with data from actual usage as represented in corpora. Closely related to the two theoretical innovations are the languagedescriptive and the cognitive-linguistic aims of the present study, i.e. the corpusbased description of ditransitive verbs in actual usage and the development of a usage-based model of speakers’ linguistic knowledge of ditransitive verbs. In the following two sections, these two objectives will be taken up in turn.
69
Note that what I call the basic form of ditransitive complementation is referred to as ‘ditransitive syntax’ by Goldberg (1992: 38).
66
Chapter 1
1.3.2 A corpus-based description of the usage of ditransitive verbs With regard to the description of the actual use of ditransitive verbs in authentic discourse contexts, I will draw on two corpora of present-day British English: the British component of the International Corpus of English (ICE-GB), which comprises 1 million words and is fully syntactically parsed, and the British National Corpus (BNC), 100 times the size of ICE-GB. The two corpora and their analysis for the purpose of the present study will be documented and discussed in detail in chapter 2. Suffice it to say at this stage that the two corpora fulfil two different functions in the present work: while ICE-GB will be analysed exhaustively for high-frequency verbs, the BNC will be taken into consideration with regard to selected low-frequency verbs and related linguistic constructions whenever the database obtained from ICE-GB proves too small. Firstly, I will take stock of the repertory of ditransitive verbs to be taken into linguistic consideration. Thus, a corpus-based list of all ditransitive verbs which meet the definition given in (11) has to be made. Secondly, all ditransitive types of complementation which are associated with the ditransitive verbs occurring in the corpus have to be categorised. This categorisation refers both to the level of lexicogrammatical patterns in which a given verb is attested (e.g. ‘V n to n’, ‘V n for n’ with the verb send) and to the level of functional categories so that, for example, correspondences between formally different patterns at the level of the linear order of clause elements (e.g. SVOdOi) can be systematised. On this basis, a quantitative analysis of individual ditransitive verbs and their patterns is necessary. As far as possible, the frequencies obtained from the quantitative analysis will have to be explained in terms of communicative principles and discourse factors which lead the language user in a given situation to prefer, say, a specific ditransitive pattern over others. For example, in 22.5% of all instances in ICE-GB, the ditransitive verb send is used in the pattern ‘V n to n’, whereas the pattern ‘V to n n’ is attested only sporadically (<1%). The close inspection of large amounts of actual language use in context will make it possible to find out reasons for this distribution of patterns; all of them are formally possible, yet they are made use of to significantly different extents. Note that the results of such a frequency-based functional analysis often call into question previously held and largely intuition-based assumptions about the use of ditransitive verbs. For example, Hudson (1992: 268) in his analysis of the double object construction states that indirect objects “typically occur between the verb and its [direct] object”. However, this hypothesis is at odds with the frequencies of the patterns of the verb send in ICE-GB.70 In fact, in 21.8% of all cases send is used in the pattern ‘V n n’, which corresponds to the seemingly ‘typical’ construction. But in another 22.5% of all cases, send is used in the pattern ‘V n to n’, in which the direct object is placed between the verb and the indirect object. At least for the verb send, it is thus questionable whether the former construction is typical, given the fact that the latter construction is just as likely to occur. 70
This verb will be discussed in detail in section 3.2.3.
Ditransitive Verbs in Previous Research and in the Present Work
67
Chapter 3, which is devoted to the quantitative and functional analysis as described and exemplified above, will offer new insights into the actual use of ditransitive verbs and, by drawing on corpus-based methods, will pay particular attention to the probabilities of ditransitive verbs and their patterns as well as to the semantic and pragmatic reasons for the choice of specific forms in given contexts. Since the present study draws on linguistic corpora designed as representative samples of language use in general, it stands to reason that from the corpus analysis general trends in language use can be extrapolated. In my view, the performance-related results obtained from the corpus analysis will increase the plausibility of cognitive models of language. As Halliday (1991: 31) points out, probabilities in language use reflect probabilities in the cognitively entrenched language system: “It had always seemed to me that the linguistic system was inherently probabilistic, and that frequency in text was the instantiation of probability in the grammar.” Note that, in two regards, Halliday’s statement is in fundamental conflict with generative approaches to language cognition (along the lines of the Chomskyan notion of competence, that is). First, Halliday obviously considers frequencies in language use to be relevant to the description of the underlying language system. Second, he does not view the language system as a set of rules which is independent of actual use. Rather, the language system, that is speakers’ linguistic knowledge, is an abstraction of the language produced and received by and around them.71 If we accept the idea that language use and language system are intricately interwoven (something which defies the rigid Chomskyan dichotomy between competence and performance), it is possible and reasonable to proceed from a fairly complete description of language use to a plausible and usage-based model of the language system. The interrelationship of language use and language system as well as the relevance of corpus analyses to both linguistic domains are visualised in Figure 1-10 below.72 As pointed out above, chapter 3 will deal with language use, i.e. it will operate at the lower level of Figure 1-10. Chapter 4, on the other hand, will be an attempt to sketch out a corpus-based and frequency-oriented usage-based model of the underlying language system with regard to ditransitive verbs. This refers to the upper level of Figure 1-10, to which I will briefly turn in the following section.
71 The Hallidayan approach to grammar as an abstraction from probabilities in language use also underlies Bybee and Hopper’s (2001a: 19) notion of a ‘stochastic grammar’: “Patterns of usage and particular choices made by speakers at any given moment are heavily influenced by both immediate and long range experience with language. Intuitions about grammaticality are based on this experience.” A more detailed discussion of the grammaticalisation framework, on which this approach rests, will be provided in section 3.3.1. 72 In a way, the use of corpus data for a usage-based model puts into operation Schmid’s (2000: 39) ‘From-corpus-to-cognition principle’. Unfortunately, Schmid himself does not go into details about the implications of this principle (cf. Esser’s (2002) critical remarks), which will be taken up in section 2.5.
68
Chapter 1
language system instantiation
usage-based model
abstraction
language use
corpus-based analysis
Figure 1-10: From a corpus-based analysis of language use to a usage-based model of the language system 1.3.3 A usage-based model of speakers’ linguistic knowledge of ditransitive verbs In the present work, a usage-based model of ditransitive verbs will be developed. In this, the concept of a usage-based model in construction grammar will be modified in two regards. Firstly, frequencies in text will be included in the cognitive model. This will allow for the distinction of different strengths of linkage between specific verbs and specific patterns and vice versa. A second modification of Langacker’s original cognitive grammar approach is the inclusion of principles and factors (e.g. principle of end-focus, context-dependent omission of semantic roles) which lead language users to prefer specific ditransitive patterns to others in particular communication situations. It appears to me that what has been largely ignored by cognitivists so far is that such ‘principles of pattern selection’ (Mukherjee 2001, 2002b) can be readily assumed to form part of speakers’ linguistic knowledge in the sense of their ‘communicative competence’ (Hymes 1972). Thus, it will be necessary to complement the lexical and constructional networks, as established in cognitive grammar, with a pattern-selectional component which includes these principles and factors and which may help explain (and visualise) why and when speakers tend to use specific verbs in specific patterns and vice versa. Note in this context that I will not confine myself to the ‘dative alternation’ between the post-verbal use of two object noun phrases on the one hand (e.g. John gave Mary the book) and the use of the prepositional variant of the indirect object (e.g. John gave the book to Mary). This binary choice (and the factors involved in this choice) has been analysed in much detail (cf. e.g. Arnold et al., 2000; Gries, 2003; Wasow and Arnold 2003), but these analyses do not take into account that ditransitive verbs also frequently occur in many other patterns. In using a verb like give, the many patterns of the verb are available to the speaker, and the present work is intended to shed light on the principles and factors that are responsible for a speaker’s choice of any of the patterns of give. The basic assumption is that these principles and factors should
Ditransitive Verbs in Previous Research and in the Present Work
69
be viewed as an integral part of speakers’ linguistic knowledge in general, i.e. their linguistic ‘competence’. By focusing on ditransitive verbs, the present study aims at opening up new perspectives in general for future corpus linguistics. Not only does corpusbased research provide insights into language in use, but it can also be a startingpoint for the modelling of the cognitive basis of language in terms of the language system which underlies language use. In principle, this procedure reflects what Chafe (1992) suggests in the following: What then is a ‘corpus linguist’? I would like to think that it is a linguist who tries to understand language, and behind language the mind, by carefully observing extensive natural samples of it and then, with insight and imagination, constructing plausible understandings that encompass and explain those observations. Anyone who is not a corpus linguist in this sense is, in my opinion, missing much that is relevant to the linguistic enterprise. (Chafe, 1992: 96) Following Chafe’s view of what corpus linguistics should be about, the present study with its consideration of corpus-linguistic theory and methodology (see chapter 2) is based on the close inspection of ‘extensive natural samples’ of language use (see chapter 3) from which a model of language ‘in the mind’ (see chapter 4) will be deduced.
Chapter 2 Methodology In this chapter, I will start off by sketching out the corpus-based approach to ditransitivity of the present study (see section 2.1). Afterwards, I will delve more deeply into the corpus material that has been used for the present study (see section 2.2). As corpus data themselves can only provide the raw material for linguistic analyses, it is necessary in this context to discuss also how linguistic theories and models may be abstracted from corpus data (see section 2.3). In the final part of this chapter, I will focus on the relation between actual corpus data and abstract linguistic models from another perspective: taking up the line of argumentation in section 1.3.3, it will be shown that corpus data may shed light on the cognitive entrenchment of language (see sections 2.4 and 2.5). 2.1
A corpus-based approach to ditransitivity
In a ‘corpus-based’ approach to ditransitivity, corpus data play an important role and are embedded in a linguistic methodology in which neither previous research into ditransitive verbs nor the need for the linguist’s careful analysis of the data and his/her informed abstraction of a linguistic model is ignored. In other words, the advent of large, computerised corpora does not at all render the linguist’s intuition and competence obsolete. Specifically, I agree with J. Aarts (1991, 2000), de Mönnink (2000) and others on the point that any linguistic description must initially be based on hypotheses that may be derived from a critical review of existing literature and/or the linguist’s own experience. In a second step, the corpus serves as a test bed for these hypotheses. Here, the tentative intuitionbased and/or experience-based hypothetical model is applied to actual performance data. Additionally, it may be useful to analyse larger corpora if the initial corpus material proves too small. From the observation of corpus data, conclusions are drawn that should ideally result in an observation-based model of language use that is able to generalise and abstract away from the versatility of performance data.1 Coming full circle, corpus evidence may, as a matter of fact, lead the linguist to refine, modify or even replace the initial working hypotheses.2 This cyclical procedure is similar to de Mönnink’s (2000) innovative ‘data cycle for descriptive linguistics’, which is intended to combine corpus data, intuitive data (e.g. informants’ intuitions on the well-formedness of linguistic forms) and 1
Note that I follow J. Aarts (1991) in distinguishing between ‘performance’ and ‘language use’ (see section 2.3). 2 This method has, of course, a long tradition in the linguistic sciences. For example, it is clearly reminiscent of Mulder’s (1968: 5) ‘hypothetical deductive method’, which in turn is based on Popper’s (1965) description of the logic of scientific discovery.
Methodology
71
the linguist’s own hypotheses. For the present study into ditransitivity, a similar data cycle can be sketched out, which is visualised in Figure 2-1.3 previous research into ditransitivity a pluralist definition of ditransitivity model of ditransitivity in language cognition
loop 2
model of ditransitivity in language use
basic cycle
corpus analysis I: ICE-GB
loop 1
corpus analysis II: BNC
conclusions from corpus analysis Figure 2-1:
A pluralist and corpus-based approach to ditransitivity in language use and cognition
The basic cycle in Figure 2-1 (proceeding from initial hypotheses over corpus data, conclusions and a model of language use to the verification or falsification of the initial hypotheses) is complemented with two additional loops. First, as mentioned above, it is sometimes necessary to analyse a second corpus that is larger than the first corpus, especially when it comes to ditransitive verbs that are used only rarely (cf. the dotted line to the right: ‘loop 1’). Second, as has been sketched out in section 1.3.3, the corpus-based model of language use can serve as a basis for the development of a model of the cognitive entrenchment of ditransitive verbs (cf. the dotted line to the left: ‘loop 2’). Figure 2-1 makes it clear that the present ‘corpus-based’ study does not stand in the tradition of ‘corpus-driven’ linguistics, as advocated, among others, by Sinclair (1991), Francis (1993), Tognini-Bonelli (1996), Hunston and Francis (2000). A detailed position statement on the aims and methods of corpus-driven linguistics (CDL) is offered by Tognini-Bonelli (2001): 3
The model shown in Figure 2-1 deviates from de Mönnink’s (2000) ‘data cycle’ in that intuitive data are not taken into consideration because the focus here is not on the range of possible linguistic forms, but rather on what (and to what extent) is actually used.
72
Chapter 2 It is accepted that there is no such thing as a theory-neutral stance, but in CDL the attempt is made to suppress all received theories, axioms and precepts ... . Obviously, as experience grows there will be new hypotheses that arise from the investigations, and if those are generally accepted they will form part of CDL methodology. Specifically in the present intellectual climate, CDL does not accept prima facie those theories, axioms and precepts that were formulated before corpus data became available. (Tognini-Bonelli, 2001: 178f.)
In effect, corpus-driven linguistics dispenses with any initial hypotheses and linguistic models. Rather, the starting point for any linguistic theory is the corpus itself. Not only is there a deep distrust of pre-corpus models, but also of linguists’ intuitions and concepts, which are believed to interfere with an unbiased analysis of the corpus data. However, this dogmatic stance with its fixation about corpus data seems both unrealistic and implausible to me. It is unrealistic because any linguistic research activity stems from some sort of initial intuitions about language. For example, a study of ditransitive verbs can hardly ever be corpusdriven because the concept of ditransitivity itself is a received idea that has a well-established tradition in linguistics (see section 1.2). Such a received idea is always at the basis of corpus-linguistic research into any linguistic phenomenon because the phenomenon at hand has to be defined (or at least circumscribed) initially before corpus data are scrutinised. So, there always is some sort of theroretical preconception involved, and, what is more, even the avoidance of an a priori theory is a theoretical preconception. The distrust of intuition in CDLmethodology is also implausible since any corpus is compiled on the grounds of linguists’ informed intuitions about language in the first place. In compiling a corpus, the following questions, among many others, have to be answered before the corpus as a data resource is available: What should the overall corpus size be? Which genres are to be covered? How should the genres be balanced? Is the corpus to be a full-text or a sample-text corpus? How, in general, can representativeness of the corpus be ensured (because it is only then that general trends in language can be extrapolated from corpus findings)? In trying to answer such questions, linguists take many intuition-based decsions on how to compile a corpus. If, however, the corpus itself is bound to be construed on these largely intuitive premises, it is implausible to disregard the usefulness (and, in fact, the inevitability) of the linguist’s intuition once the corpus has been compiled.4
4
Interestingly enough, Tognini-Bonelli (2001: 178) herself mentions the impossibility of having a ‘theory-neutral stance’ (see above). This caveat should not only apply to language-specific theory but also to corpus-related theory. For example, I would contend that even a CDL-linguist would automatically turn to a larger corpus if he/she did not find a specific linguistic form in a smaller corpus. The trivial decision to consult a larger corpus would itself be based on a preconception about the representativeness of corpora which is not based on corpus data as such.
Methodology 2.2
73
Corpus data and corpus analysis
In this section, I would like to present the corpora that have been used for the present study. Special emphasis will be placed on ICE-GB (see section 2.2.1). In particular, principles and problems of the parsing scheme of ICE-GB will be discussed (see section 2.2.2). These problems notwithstanding, a list of all the ditransitive verbs that match the definition of ditransitive verbs suggested in section 1.3.1 can be obtained (see section 2.2.3). Finally, I will touch on the BNC, which serves as an ancillary to ICE-GB in the present work (see section 2.2.4). 2.2.1 The International Corpus of English – the British Component (ICE-GB) The design of the International Corpus of English (ICE) goes back to Sidney Greenbaum, who launched the project in the late 1980s (cf. Greenbaum, 1996). In its final version, the corpus will include some twenty subcorpora of English, each representing a regional variety of English and comprising 1 million words of spoken and written English. Not only native varieties of English (e.g. American, Australian and British English) but also non-native varieties of English as a Second Language (e.g. Hong Kong, Indian and Singaporean English) are covered. All 1-million-word subcorpora are compiled according to the same standards. They include 500 text samples from 1990-1993, each 2000 words in length. One particular advantage of ICE corpora (in comparison with, say, other 1-millionword corpora of the Brown corpus type) lies in the fact that spoken usage makes up for 60% of the corpus. Thus, the priority that speech is usually given over writing in modern linguistics (cf. Lyons, 1981: 11ff.) is now accommodated by an appropriate corpus design. The text categories which are included in all ICE subcorpora are listed in Tables 2-1 (for the spoken component) and 2-2 (for the written component).5 The British component of ICE (ICE-GB) was published in 1998. ICE-GB has been fully tagged and parsed. The parsing scheme follows, by and large, the grammatical categories at clause, phrase and word level that can be found in the Comprehensive Grammar by Quirk et al. (1985).6 A detailed manual to ICE-GB is also available (cf. Nelson et al., 2002), which includes all the information on corpus design, sources of corpus texts, software facilities, principles of tagging and parsing and related issues. In addition, the exact number of words that each spoken and written genre and sub-genre includes is given in the Appendix; the figures have been obtained with the corpus retrieval software (ICECUP 3.0, 1998: s.v. corpus map). 5
Many principal distinctions in the ICE corpora go back to the design of the original Survey of English Usage (SEU) Corpus (cf. Quirk, 1968: chapter 7). 6 It should not go unmentioned that the parsing of ICE-GB is to a large extent based on the TOSCA scheme that has been developed by Jan Aarts and other linguists at the University of Nijmegen. An overview of their classification scheme is provided by J. Aarts et al. (1996).
74
Chapter 2
Table 2-1:
Spoken text categories in ICE (Nelson et al., 2002: 307)
Table 2-2:
Written text categories in ICE (Nelson et al., 2002: 308)
Methodology
75
The corpus itself is linked to the software tools of the ICE Corpus Utility Program (ICECUP). ICECUP gives access to the wealth of parsing information in ICE-GB. Of particular importance are the Fuzzy Tree Fragments (FTFs) that can be designed by using ICECUP (cf. Nelson, 1998; Wallis, 1998; B. Aarts, 2000; Nelson et al., 2002). These FTFs make possible exhaustive searches for specific syntactic structures (with or without a definition of particular word forms). An FTF is made up of three sectors: ‘function’, ‘category’ and ‘feature(s)’. Since all clauses (technically representing ‘parse units’) are also heavily annotated for verbal and clausal characteristics (e.g. aspect, fronting, transitivity), it is thus possible to automatically retrieve all parse units in which ditransitive verbs are used. The search for instances of ditransitivity may be restricted to specific verbs, e.g. the word form give of the lemma GIVE. The FTF for this search query is given in Figure 2-2.7
Figure 2-2:
An FTF-based search query in ICE-GB
The search query in Figure 2-2 gives all the instances of give (i.e. the word-form specified in the lower part) in which give (1) functions as a verb (cf. ‘VB’ in the 7
It should be noted that more complex FTFs can also be designed. Complex FTFs are made up of two or more nodes. It is thus possible to look for a sequence of syntactic functions (in general or in specific formal realisations). Also, brief mention should be made of genre-specific search options. Searches can thus either be carried out across the corpus in its entirety or be confined to specific genres.
76
Chapter 2
top-left box) and (2) forms part of a verb phrase (cf. ‘VP’ in the top-right box) and (3) in which it is parsed as ditransitive (cf. ‘ditr’ in the lower box) in ICE-GB. For any occurrence, it is then possible to obtain the entire parsing information in tree format. One example is given in Figure 2-3: I’ll give you the school’s number . As Figure 2-3 illustrates, each element at word, phrase and clause level is analysed from three perspectives: (1) its function (specified in the top-left box of each node); (2) its formal category (specified in the top-right box of each node); (3) its additional features (specified in the bottom sector of each node). Thus, the word give, for example, is analysed in this sentence as the main verb (‘MVB’ = function), as a verb (‘V’ = category) and as ditransitive (‘ditr’ = feature). Similar tripartite analyses are carried out for all other words, but also for all clause elements (e.g. the indirect object ‘OI’ realised as a noun phrase ‘NP’) and the clause as such (i.e. the parse unit ‘PU’ realised as a ‘main’ clause ‘CL’).8
Figure 2-4:
An example of parsing in ICE-GB
From what has been said so far about ICE-GB, three main reasons can be given as to why the present study has been based on this particular corpus. First, it represents spoken and written English from the early 1990s. Second, it is fully tagged 8
The design and development of the Survey Parser, used for ICE-GB, is described in detail by Fang (1996).
Methodology
77
and parsed, which makes it possible to draw on detailed morphosyntactic annotations for some automatic search queries. Third, it is of a reasonable size which still allows for a manual analysis of the data. As Kennedy (1998: 68) correctly points out, “there is no point in having bigger and bigger corpora if you cannot work with the output.” 2.2.2 Parsing of ditransitive verbs in ICE-GB: principles and problems The parsing scheme in ICECUP distinguishes between the following seven verb complementation types (cf. ICECUP 3.0, 1998: s.v. The ICE Tagging Scheme > Verb (tag) > special cases > complementation > category):9 a) b) c)
d)
e)
f)
g)
9
Intransitive (intr) Intransitive verbs are not followed by any object or complement: She works in a library. She graduated last summer. Stop. Copular (cop) Copular verbs require the presence of a subject complement: It really is great fun. It isn’t my fault. She ’s home. ... Monotransitive (montr) Monotransitive verbs are complemented by a Direct Object only: I buy too many books. She left her coat in the cinema. Please sign both sides of the form. [...] Dimonotransitive (dimontr) Dimonotransitive verbs are complemented by an Indirect Object only. They include show, ask, assure, grant, inform, promise, reassure, and tell: When I asked her, she burst into tears. I’ll tell you tomorrow. Show me. Ditransitive (ditr) Ditransitive verbs are complemented by an Indirect Object and a Direct Object: We tell each other everything. They built themselves a new theatre. Give us some more drinks. Complex transitive (cxtr) Complex transitive verbs (cxtr) are complemented by a Direct Object and an Object Complement: He knocked himself unconscious. It would make me sick. I take that as a compliment. ... Transitive (trans) The transitivity of a verb may be unclear in some instances if it is complemented by a nonfinite clause: I expect John to do the hoovering.
This classification scheme, including definitions and examples, is obtained from the online manual to the retrieval software ICECUP 3.0 (1998); the same definitions and examples can also be found in the ICE-GB handbook (Nelson et al., 2002: 38ff.). Note also that the annotation of verbs according to their complementation types is subsumed under grammatical tagging in ICECUP, while I prefer to regard it as a parsing feature because the annotation is derived from the analysis of all the clause elements in the parse unit at hand.
78
Chapter 2 Here the noun phrase John may be the Subject of the nonfinite clause or the Object of the host clause. In all such cases we tag the main verb as V(trans,...). ... However, the trans label is not applied (a) if the verb is be: One of my aims is to finish my PhD. V(cop,pres) All he did was repeat what someone else has told him. V(cop,past) (b) if the nonfinite clause does not have an overt Subject: She helped design the course. V(montr,past) I enjoy doing it. V(montr,pres) (c) if the noun phrase is followed by a wh-clause whose main verb is a toinfinitive: No one can tell me what to do. V(ditr,infin) The manual shows you how to install it. V(ditr,pres)
This overview of the verb complementation types already illustrates the main principles of the transitivity annotation and, at the same time, some of the problems they pose for the present study. As for the principles of parsing, the distinction of the seven verb transitivity types is entirely syntactic in nature.10 In other words, firstly, the transitivity of a verb is determined on the grounds of all the clause elements that are required by the verb and that are attested in a given clause. Thus, it is not surprising that sentences such as Show me are not parsed as ditransitive because the second object is not explicitly stated. Secondly, the purely syntactic identification of transitivity types implies that one and the same verb is parsed differently in different sentences, because transitivity is not seen as a more or less stable lexical property of the verb itself. Thus, the verb show in Show me is parsed as dimonotransitive, whereas it is parsed as ditransitive in The manual shows you how to install it. Focusing on ditransitive verbs, this syntactically motivated parsing scheme is to a large extent at odds with the primarily lexical and semantic definition of ditransitivity that underpins the present study (see section 1.3.1) in which Show me and The manual shows you how to install it are taken to belong to one and the same transitivity category. Also, not all instances of the verbs that are covered by my understanding of ditransitivity are parsed as ditransitive in ICE-GB. For example, the following instances of give are classified as monotransitive (montr): (12) (13) (14)
Alice and I have given (montr) quite a few I don’t know whether he was giving (montr) it Oh just give (montr) a ring
Given the definition of ditransitive verb complementation on which the Survey Parser draws (see above), it does not come as a surprise that whenever the affected entity is not realised as an indirect object at the level of syntactic surface structure, the verb is not parsed as ‘ditransitive’ but as ‘monotransitive’ (or as ‘dimonotransitive’ whenever the transferred entity is not stated in the form of a 10 It should be noted in passing that Fang (1996: 145), in describing the Survey Parser, lists another transitivity type: ‘complex-ditransitive (V + Indirect Object + Direct Object + Object Complement)’ as in I sold him my car secondhand. This type, however, has obviously not been included in the final version of ICECUP.
Methodology
79
direct object). Accordingly, the label ‘intransitive’ is assigned whenever no object is detectable in the parse unit. In a sense, then, ditransitivity in ICECUP could be regarded as based on what I wish to call ‘explicit ditransitive syntax’: both objects are made explicit, with the indirect object realised as a noun phrase and the direct object as a noun phrase or a clause. That give could still be seen as semantically ditransitive in cases where, for example, the affected entity of the eventtype of GIVING is omitted (e.g. because it is contextually reconstructible) is not taken into account by the Survey Parser. Also, the explicit ditransitive syntax to which ICECUP confines itself does not consistently cover patterns in which the indirect object (representing the affected entity) is realised as a to-phrase which follows the direct object. What is more, such instances are sometimes parsed as monotransitive (‘montr’), as in (15), and in other cases as complex-transitive (‘cxtr’), as in (16): (15) (16)
... for pressure to be brought to bear on BS to sell (montr) its Scottish assets to any potential buyers I think if you send (cxtr) that to him he’s got to address the issue this time certainly
On the other hand, the label ditransitive is used for some instances of verbs which are not complemented by two objects realised as two noun phrases. In particular, complementations of verbs with a noun phrase (functioning as the indirect object) followed by a wh-clause or a that-clause (functioning as the direct object) are also parsed as ditransitive (‘ditr’): (17) (18) (19)
And then I showed (ditr) him what was wrong with the other one However, no one can tell (ditr) me what to do about it Tell (ditr) him we are waiting for the order
In conclusion, the examples mentioned so far show that there are clear differences between the comparatively narrow definition of ditransitivity in ICECUP (and its parsing scheme) and the broad and semantically-oriented definition of the present study. Thus, it is not possible to rely on the parsing information included in ICEGB in order to automatically find all instances of ditransitivity in the corpus. In a wider setting, this mismatch illustrates the potential problems that are involved in working with heavily annotated corpora. On the one hand, it is very useful to have access to corpus data that are not only tagged but also fully parsed, because this may facilitate search queries and help avoid manual analyses when it comes to the realm of syntax. On the other hand, the utilisation of parsing information always presupposes that the linguist keeps in line with the definitions and concepts that are at the basis of the parsing scheme at hand. There is a danger, therefore, that already available corpora with their syntactic annotation predetermine the linguistic theory of and research into syntax. As a matter of fact, the reverse
80
Chapter 2
order should be aimed at: not the corpus annotation should influence linguistic research, but linguistic research questions should be the guideline for the corpus annotation. This, of course, is reminiscent of what Sinclair (1991: 21) calls a ‘clean text policy’: “The safest policy is to keep the text as it is, unprocessed and clean of any other codes. These can be added for particular investigations.” If parsing information, for example, becomes an integral part of the corpus, the practical advantages of their availability may soon be outweighed by the unpleasant side-effect that the parsing information itself narrows down the range of possible research foci. 2.2.3 A corpus-based list of ditransitive verbs Although it is not possible to automatically retrieve all instances of verbs (as tokens) that are covered by the definition of ditransitivity in (11), the narrow understanding of ditransitivity (in terms of an explicit ditransitive syntax) which underpins the parsing scheme of ICE-GB can be utilised for an exhaustive identification of all the verbs (as types) that occur in the corpus at least once in an explicit ditransitive syntax and that should also be considered ditransitive from a pluralist point of view. For convenience, the definition of ditransitive verbs given in (11) on p. 65 is repeated here: (11)
A working definition of ditransitive verbs A ditransitive verb (DV) is a trivalent verb that requires a subject (S), a direct object (Od) and an indirect object (Oi) for a complete syntactic complementation. It is necessary for all clause elements to be realisable as noun phrases (NPs): this realisation (S:NP – DV – Oi:NP – Od:NP) is called the basic form of ditransitive complementation. If a verb is attested in the basic form of ditransitive complementation in actual language use, it is also considered a ditransitive verb in all other forms of complementation. All ditransitive verbs and ditransitive complementations are associated with an underlying proposition that represents the situation type of TRANSFER with three semantic roles involved: the ditransitive verb denotes an action in which the acting entity transfers a transferred entity to the affected entity.
Since this definition, although semantically-oriented, includes a formal aspect, it is possible to make use of ICECUP in order to identify all the verbs that occur in ICE-GB at least once in the basic form of ditransitive complementation (S:NP – DV – Oi:NP – Od:NP). In a first step, then, a search for all instances of a verbal clause element (VB), realised as a verb phrase (VP) with a ditransitive main verb (ditr), can be taken to include all verbs that meet the formal criterion of ditransitive verbs above. As shown in Table 2-3 below, this search query results in 1820 matches. However, 79 cases of these matches are questionable when it comes to the occurrence of verbs in the basic form of ditransitive complementation. This pertains to those verbs in particular which are parsed as ditransitive because they
Methodology Table 2-3: Search query
81 Identifying ditransitive verbs in ICE-GB by using ICECUP
Fuzzy Tree Fragment verbal (VB) verb phrase (VP) ditransitive (ditr)
Matches (automatic) 1820
Clear matches (manual) 1741
are complemented by a noun phrase and a wh-clause or that-clause but which cannot occur in the basic form of ditransitive complementation with both objects realised as noun phrases (e.g. advise, inform, remind). These cases have been excluded by manually checking the data. Thus, we are left with 1741 clear matches, i.e. examples of verbs that may occur in the basic form of ditransitive complementation. In effect, all verbs in the 1741 examples are ditransitive verbs according to the formal criterion included in the pluralist definition of ditransitivity in (11). From the 1741 clear matches, a corpus-based list of ditransitive verbs (as types) can be obtained. This list is given in Table 2-4. Those verbs that are given in boldface in Table 2-4 are particularly frequent and can thus be regarded as ‘typical’ and ‘habitual’ ditransitive verbs. The frequency of all verbs in all ditransitive patterns can only be specified by manually searching for each of the verbs individually in the corpus. As has been discussed in section 2.2.2, examples of ditransitivity (as defined in the present study) are often not parsed as ditransitive in ICE-GB, so that the 1741 examples, on which Table 2-3 is based, do not provide an exhaustive list of all the instances that are covered by the working definition of ditransitivity in (11). It is therefore important to note that the frequencies in Table 2-4 refer to the ICECUP-generated – and manually postedited – list of those 1741 cases that are parsed as ditransitive in ICE-GB. Nevertheless, the corpus-based list of ditransitive verbs in Table 2-4 fulfils two important functions: (1) it is a list of all the verbs (as types) that occur in ICE-GB and are ditransitive in the sense of the present study (although not all tokens are covered); (2) as will be discussed below, the frequency results in the list are an appropriate means for categorising ditransitive verbs into groups of typicality. The extent to which a specific verb is typical of a given verb class is relevant to both the description of language use and the modelling of language cognition. Typical ditransitive verbs can be taken to predetermine the usage of novel verbs (e.g. fax, email) in ditransitive patterns, i.e. typical ditransitive verbs provide guidelines for what Goldberg (1995: 120ff.) calls the ‘partial productivity’ of constructions. Also, typical ditransitive verbs are cognitively entrenched as prototypical entries around which less typical, but semantically similar verbs orbit, as it were. In order to identify those verbs that are typical of the class of ditransitive verbs (for example, when it comes to the classification of novel ditransitive verbs), frequency is of paramount importance:11 11
Although Goldberg (1995) does not ignore the relevance of frequency in this context, it is somewhat amazing, given the issue’s particular importance for models of language cognition, that she mentions it only in passing.
82
Chapter 2
Table 2-4:
A list of all the verbs in ICE-GB (as types) that occur in the basic form of ditransitive complementation (S:NP – DV – Oi:NP – Od:NP)
verb
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36.
accord afford allocate allow ask assign assure award bet bring buy call cause charge cook cost cut deal deliver deny design do draw drop earn feed file find fine get give grant guarantee hand keep leave
parsed as ditransitive in ICE-GB (instances) 3 4 4 19 91 4 13 7 1 7 12 1 9 5 3 23 2 1 1 8 2 27 1 4 8 1 1 2 2 34 562 8 7 5 3 8
verb
37. lend 38. lose 39. make 40. offer 41. order 42. owe 43. pass 44. pay/overpay 45. permit 46. play 47. prescribe 48. profit 49. promise 50. purchase 51. quote 52. read 53. refuse 54. render 55. save 56. sell 57. send 58. serve 59. set 60. show 61. spare 62. supply 63. take 64. teach 65. tell 66. throw 67. vote 68. win 69. wish 70. write cases of explicit ditransitive syntax
parsed as ditransitive in ICE-GB (instances) 12 2 3 54 2 8 2 19 1 2 1 1 12 1 2 1 3 1 4 1 79 1 3 84 3 1 12 23 491 2 1 2 9 5 1741
Methodology
83
On this view, frequency is expected to affect the classification of new verbs. Two kinds of frequency information need to be distinguished. On the one hand there is token frequency, which refers to the number of times a given instance (e.g., a particular word) is used in a particular construction; on the other hand there is type frequency, which refers to the number of distinct words that occur in a particular construction. (Goldberg, 1995: 134) As a matter of fact, Goldberg is especially interested in the extent to which one particular construction is preferred over others in providing the usage pattern for verbs newly introduced into a language.12 For the present study, which focuses on ditransitive verbs and patterns only, the two kinds of type and token frequency that ought to be taken into account for establishing a gradient of prototypicality of ditransitive verbs have to be modified. I suggest that the following dimensions be taken into consideration in empirically defining the frequency-based typicality of ditransitive verbs in ICE-GB: (1) the overall frequency of a ditransitive verb in the corpus; (2) the frequency with which a ditransitive verb occurs in an explicit ditransitive syntax (as can be obtained from the parsing information in ICE-GB, see Table 2-4). The two dimensions are in a way correlated since a typical ditransitive verb has to occur frequently in general and frequently in an explicit ditransitive syntax in particular. It is obvious that a verb which is only sporadically attested in language use cannot serve as a typical example of the verb class of which it is a member. Similarly, even a comparatively frequent verb can only be typical of the ditransitive verb class if a substantial part of its occurrences is associated with a syntax that makes explicit all argument roles of the ditransitive event type (as subject, indirect object and direct object respectively). For it is only against this background that, for example, language users may be expected to recover specific argument roles from the context that are not explicitly mentioned. Both kinds of frequency information can be automatically retrieved from ICE-GB by using ICECUP. They are given in Table 2-5.13 Given the figures in Table 2-5, I suggest that three groups of ditransitive verbs be distinguished on the grounds of two sorts of frequency information: (1) typical ditransitive verbs, which are used very frequently in general and also frequently in an explicit ditransitive syntax (give, tell); (2) habitual ditransitive verbs, which are used fairly frequently in general but not in an explicit ditransitive syntax in the clear majority of all cases in which they occur (ask, send, show, offer); (3) peripheral ditransitive verbs, which are used 12
This is related to the fact that Goldberg’s (1995) starting-point for a model of language cognition is a particular construction. In contrast, I would argue that specific verbs are the basic units, which are associated with one particular construction at the cognitive level, but with different instantiating patterns at the level of syntax (see section 1.2.7). 13 Note that while the overall frequency can be directly obtained from the results of an ICECUP-generated word count, the percentage of explicit ditransitive syntax can only be determined by manually analysing all occurrences of all the ditransitive verbs at hand.
84
Chapter 2
Table 2-5:
Overall frequency of ditransitive verbs in ICE-GB and their relative frequency in an explicit ditransitive syntax
‘typical’ ditransitive verbs ‘habitual’ ditransitive verbs ‘peripheral’ ditransitive verbs
ditransitive verb give tell show ask send offer
overall frequency 1160 794 639 518 346 198
(other verbs)
< 200 each
explicit ditransitive syntax 562 48.4 % 491 61.8 % 84 13.0 % 91 17.6 % 79 22.8 % 54 27.3 % OR
< 5.0 % each
only sporadically in general and/or which are used only rarely in an explicit ditransitive syntax (all remaining verbs listed in Table 2-4, p. 82). In Figure 2-4, the three distinct clusters that emerge from the two kinds of frequency information are visualised graphically. overall frequency
give
1000 typical ditransitive verbs tell
750 show 500
ask habitual ditransitive verbs send
250
offer peripheral ditransitive verbs (64 verbs) 25%
Figure 2-4:
50%
75%
percentage of explicit ditransitive syntax
A frequency-based distinction of typical, habitual and peripheral ditransitive verbs in ICE-GB
Methodology
85
The distinction between the three groups of ditransitive verbs implies that special emphasis will be placed on typical and habitual ditransitive verbs in language use (see sections 3.1 and 3.2) and language cognition (see chapter 4). This will be based on an exhaustive analysis of all occurrences of the following six verbs in ICE-GB: give (1160 instances), tell (794 instances), show (639 instances), ask (518 instances), send (346 instances), offer (198 instances). For peripheral ditransitive verbs, for which ICE-GB turns out to be too small a database, selected data from the BNC will be used (see section 3.3). Finally, note that the formal criterion that is included in the present definition of ditransitivity excludes all those verbs from further investigation that do not occur in the basic form of ditransitive complementation but whose semantics may resemble the typically ditransitive situation type of TRANSFER. As pointed out in section 1.2.1, the verb provide is a case in point since its semantics is very similar to give. However, since in the corpus data at hand it does not occur in the basic form of complementation, it is not considered a ditransitive verb on formal grounds. In the final resort, then, the present study of ditransitivity deals with a formally defined subset of all the verbs that may be regarded as ditransitive from a purely semantic perspective. 2.2.4 The British National Corpus (BNC) In the present study, the BNC serves as an ancillary corpus for the description of peripheral ditransitive verbs in ICE-GB. What will be used is the second version of the BNC, i.e. the BNC World Edition published in 2000.14 The BNC comprises some 100 million words from roughly 4,000 texts. Broadly, 90% of the corpus are written sources, 10% are spoken texts. The written texts cover non-fictional genres (informative prose from 1975 to the early 1990s) and fictional writing (imaginative from 1960 to the early 1990s). The corpus is fully tagged (following two different tagging schemes), but does not include any parsing information. Its enormous size makes it possible to analyse low-frequency items (i.e. peripheral ditransitive verbs) in much more detail than with the help of standard-size corpora of 1 million words such as ICE-GB. 2.3
From corpus data to linguistic theory
As visualised in Figure 2-1 (p. 71), the present study aims at two different kinds of models of ditransitivity to be derived from corpus data: (1) a model of ditransitivity in language use; (2) a model of ditransitivity in language cognition. While I will delve more deeply into the latter aspect in section 2.4, this section is devoted to the former aspect, i.e. the corpus-based description of language use. In 14
A printed handbook with detailed information on the corpus design and the standard software SARA is available for the first version (cf. Aston and Burnard, 1998). For the BNC World Edition, a detailed manual is available online (cf. Burnard, 2000).
86
Chapter 2
particular, two theoretical questions that arise from modern corpus-linguistic methodology have to be tackled: (a) What is the status of corpus linguistics? (b) What is the relation between corpus data in all their complexity and versatility on the one hand and models of language use as necessarily idealised accounts on the other? Without any doubt, modern corpus-linguistic research has led to an entirely new perspective on language (cf. Lenz, 2000: 12). Yet, even corpus linguists themselves continue to disagree on the status of corpus linguistics. On the one hand, some argue that despite the wealth of new corpus-based insights into actual language use, corpus linguistics should be seen as nothing more than a modern computer-aided linguistic methodology (cf. Leech, 1992: 105; Kennedy, 1998: 268; B. Aarts, 2000: 7; Meyer, 2002: xi ff.). On the other hand, we also find the view that corpus linguistics should be considered a discipline in its own right because many aspects of language use can only be described by drawing on corpus data, especially when it comes to the multitude of linguistic routines such as collocations, colligations and preferred lexicogrammatical patterns: thus, Teubert (2001: 125) states that “European corpus linguistics is gradually becoming a sub-discipline in its own right”, and Tognini-Bonelli (2001: 177) explicitly speaks of a “new discipline within linguistics”. As I have argued elsewhere (cf. Mukherjee, 2002a: 51ff.), the two points of view are not mutually exclusive. I would contend that corpus linguistics represents both a new method (in terms of computer-aided descriptive linguistics) and a new discipline (in terms of a new approach to language description). In a sense, an analogy can be drawn between corpus linguistics and other scientific fields in which the development of new methods have gradually led to new insights and to the establishment of a new discipline. A good case in point is the coming of age of microbiology. In their standard introduction to microbiology, Madigan et al. (1997), for example, point out that the development of new methods (i.e. microscopes in particular) and of an independent branch within biology (i.e. microbiology) are inseparable from a historical point of view: Although the existence of creatures too small to be seen with the eye had long been suspected, their discovery was linked to the invention of the microscope. ... Leeuwenhoek’s microscopes were extremely crude by today’s standards, but by careful manipulation and focusing he was able to see organisms as small as prokaryotes. ... Only in the nineteenth century did improved microscopes become available and widely distributed. During its history, the science of microbiology has traditionally taken the greatest steps forward when better microscopes have been developed, for these enabled scientists to penetrate ever deeper into the mysteries of the cell. Microbiology as a science did not develop until the latter part of the nineteenth century. This long delay occurred because, in addition to microscopy, certain basic techniques for the study of microorganisms needed to be devised. (Madigan et al., 1997: 20f.)
Methodology
87
As for corpus linguistics, it seems to me that there is a similar correlation between its methodology and its status as a discipline in its own right.15 Specifically, the compilation of large and representative corpora as well as the development of relevant software tools exemplify methods which have allowed for entirely new insights into language use. In a wider setting, these corpus-based insights have had such wide-ranging effects (e.g. because many previously held assumptions about language were up for re-interpretation) that corpus-based research could be seen as a new approach to language indeed. For example, the interdependence of lexis and grammar (in terms of lexicogrammatical patterns, see section 1.2.6), which is characteristic of natural language use, can only be described by looking closely at large amounts of language data. In the final analysis, then, it is only by analysing corpora that probabilities in language use are accessible. Thus, it is obvious that corpus-based methodology leads to linguistic theories that include features and aspects that could not have been covered by linguistic research in the pre-corpus era. If we accept the fact that important aspects of language use are accessible only through corpus data (and that corpus-based linguistic models are therefore, in a way, superior to other theoretical frameworks), the question arises as to how the raw material of corpus data and linguistic models derived from corpus data are related to each other. In the field of grammar, a very useful distinction is made by J. Aarts (1991), who distinguishes between ‘performance’ and ‘language use’. Generally speaking, corpora contain a vast amount of performance data, i.e. the linguistic behaviour of many native speakers in many different communication situations. However, corpus-based models of grammar should not attempt to explain all performance data in their entirety, because the data will always include instances of, say, unacceptable language use, clear mistakes and intended ungrammaticality. It stands to reason that performances of such and similarly dubious kinds should not be included in a corpus-based grammar of language use: ... a grammar describing language use cannot be written by trying to account for each and every fact of performance; language use and performance are two distinct notions. (J. Aarts, 1991: 59) To be more precise, there exists some sort of ‘correctness filter’ (J. Aarts 2000: 29) between performance and language use, which is based on accepted norms and routines in a given speech community. While everything included in a corpus is ‘performance’, ‘language use’ refers to a descriptive level at which a first abstraction from performance data takes place. In a way, there are two factors that are responsible for filtering out from performance data those aspects of usage that ought to be included in a corpus-based model of language use: ‘frequency’ and ‘normalcy’, the latter representing the general acceptance of a linguistic form in a 15
A similar analogy between corpus linguistics and microbiology is drawn by J. Aarts (2000: 17).
88
Chapter 2
given speech community. Since at the level of language use, acceptability judgements (in terms of speakers’ attitudes, linguistic prescriptions and generally accepted norms) come into play, J. Aarts (1991) ascribes an inevitably normative quality to any model of language use: In that respect, a grammar of language use might well be a normative grammar, if we understand ‘normative’ as ‘based on the norms set by a not insignificant part of the language community’ – that is, a grammar of structures used (frequency) as well as accepted (normalcy) by a large number of language users. (J. Aarts, 1991: 58) A concept that is related to – and can be fruitfully combined with – J. Aarts’s (1991) normative and corpus-based model of language use is the notion of ‘abstracted corpus norm’ as suggested by Esser (1993). He introduces the term in order to describe Gibson’s (1966) style machine, which is intended to compare the styles of individual texts. To this end, Gibson (1966) first of all abstracts from the entirety of all texts a statistical reference norm with which relevant linguistic features of the individual texts are to be compared. It is important to note that the linguistic features under scrutiny are identified a priori and intuitively, which is reminiscent of the intuitive genre distinctions that are at the basis of the compilation of modern representative corpora: What Gibson has done, in effect, is to establish an abstracted corpus norm on the basis of independently classified texts. One may have objections to the impressionistic primary classification. But the method as such is not invalidated because it would be possible to use, as input for a comparable abstracted norm, external criteria of text selection and classification such as for example the genres in text corpora ... . (Esser, 1993: 88) In effect, the abstracted corpus norm provides a tertium comparationis to characterise and compare individual texts. While Esser (1993) uses the term ‘abstracted corpus norm’ to refer to the stylistic comparison of texts on the grounds of a statistical reference norm which is generalised from the entirety of a text corpus, I would contend that this concept can be applied to any model of language use. In principle, all corpus-based dictionaries and grammars are based on such an abstracted corpus norm, because they do not include everything that is performed but provide information on what is frequently used and/or what is considered normal usage. As shown in Figure 2-5, the abstracted corpus norm can be regarded as an operationalisation of J. Aarts’s (1991) normative understanding of language use. Corpora such as ICE-GB and the BNC are representative samples of native-like performance. What is more, corpora include performance data of many native speakers (NS), covering a wide range of different levels of proficiency, accents, dialects and social/educational backgrounds, as well as per-
Methodology
89
formance data from many different spoken and written genres. It is therefore possible to extrapolate from the entirety of performance data a model of language use by describing an abstracted corpus norm on the basis of corpus data. In other words, a model of language use which is based on an abstracted corpus norm is a supraindividual abstraction of what is frequent and/or normal in a given speech community. abstracted corpus norm (=> model of language use)
general corpus (e.g. ICE-GB, BNC)
spoken component: Genre 1: Text 1, Text 2, Text … Genre 2: Text 1, Text 2, Text … Genre …
NS 1, NS 2, NS 3, NS 4, NS 5, NS 6, NS 7, NS 8, NS 9, NS … - various levels of proficiency - various accents and dialects - various educational and social backgrounds -…
written component: Genre 1: Text 1, Text 2, Text … Genre 2: Text 1, Text 2, Text … Genre …
corpus representative of
native-like performance Figure 2-5:
The abstracted corpus norm: from performance data in corpora to a model of language use
For English ditransitive verbs, major aspects of such a model of language use will be described in chapter 3. The corpus-based model will be such that it abstracts away from actual performance what is used frequently and considered normal in (British) English. Particular attention will be paid to frequencies of ditransitive verbs and patterns in the corpus data, so that the norm in native-speakers’ linguistic behaviour can be identified. In describing this abstracted corpus norm, both language-external and language-internal factors that can be held responsible for
90
Chapter 2
this norm will be identified. Thus, the model of ditransitive verbs in language use will be based on the combination of a quantitative analysis of performance data and a qualitative analysis of general trends in terms of language use. 2.4
‘Competence’ in the light of corpus data
What may a corpus-based model of language use tell us about speakers’ internalised knowledge of the underlying language system? At first sight – and in generative terminology – this question may seem to be an attempt to relate language in performance to linguistic ‘competence’. However, a model of the language system that is based on corpus evidence does not have much in common with a generative model of competence. Thus, it is not very useful to take over and extend or redefine the term ‘competence’ as such, which would automatically lead to terminological confusion (cf. D.S. Taylor, 1988).16 Rather, I prefer to speak of a corpus-based model of speakers’ linguistic knowledge along the lines of Hymes’s (1972, 1992) concept of ‘communicative competence’. There are, at least, three fundamental differences between the generative approach to competence and a corpus-based model of speakers’ linguistic knowledge. First, generative grammar focuses on the knowledge of what is possible in language. The range of what is possible is mainly identified on the basis of intuition-based grammaticality judgements. Second, the focus is on an ideal speakerhearer. Third, as Chomsky himself has repeatedly pointed out, frequencies in text are considered irrelevant to competence, i.e. the internalised knowledge of grammar.17 On the other hand, a corpus-based model of speakers’ internalised linguistic knowledge would ideally be based on language used by real speakers in authentic contexts. It takes into account frequencies in text because the model is also intended to mirror speakers’ anticipations of what is probable. The knowledge of linguistic routines and patterns includes the ability to use them, i.e. the knowledge of principles and factors which are responsible for such routines and patterns. This ability for use corresponds to what Chomsky (1980: 224) himself describes as ‘pragmatic competence’, which he, however, clearly separates from competence proper (i.e. ‘grammatical competence’). In contrast, a corpus-based model of speakers’ linguistic knowledge would not differentiate between grammatical and pragmatic competence. More important, it is an attempt to bridge the gap between what speakers know and what speakers use. As noted in section 1.3.2, I follow in this regard Halliday’s (1991: 31) view that system and use are inseparable, that in fact language use is an instantiation of the language system. A corpus-based model of speakers’ linguistic knowledge should thus be able to
16
A more recent example of the deep distrust of corpus evidence among cognitive linguists in the generative camp is provided by Borsley and Ingham (2002). In his reply to Borsley and Ingham, Stubbs (2002) makes it clear that their highly polemical rebuttal of corpus linguistics is based on severe misunderstandings. 17 See the interview B. Aarts (2000: 5ff.) conducted with Chomsky.
Methodology
91
account for the following characteristic features of language use as accessible through corpora: • First, some linguistic forms are more frequent than others in language use, and some formally possible forms are unlikely to occur at all. It is reasonable to assume that speakers implicitly know about such probabilities of linguistic forms and their combinations. • Second, we find linguistic routines and patterns of different kinds (e.g. preferred combinations of specific words and particular constructions) so that speakers’ linguistic knowledge not only allows for infinite use but is based on routine as well. • Third, quantitative data on the frequencies and patterns in text can usually be explained by functional and context-dependent principles and factors. These principles and factors then seem to be part of speakers’ linguistic knowledge: language encoders are obviously guided by such principles and factors to make appropriate use of their linguistic means and to adhere in their linguistic behaviour to regular expectations. This observation should translate into a model which ascribes to whatever is frequent in language use a status that is different from whatever is rarely used. • Fourth, lexical and grammatical choices are interdependent in language use. The all-pervading nature of colligations and lexicogrammatical patterns calls into question the autonomy of syntax. Lexicogrammatical patterns should thus be at the basis of a model of speakers’ linguistic knowledge that is supposed to account for actual usage. 2.5
Bridging the gap: from corpus to cognition
According to the guidelines suggested in the previous section, the corpus-based model of speakers’ linguistic knowledge of ditransitive verbs will be dataoriented and frequency-based, functionalist and lexicogrammatical in nature. In modelling linguistic knowledge on the basis of quantitative data obtained from corpora, the model will be truly usage-based (see section 1.2.7). The general methodology of the present study can thus be summarised as a from-corpus-tocognition approach. This methodology is applied to ditransitive verbs in the present study, but can be expected to be applicable to all aspects relevant to the description of language use and the modelling of language cognition. This fromcorpus-to-cognition approach capitalises on what Schmid (2000: 39) calls the ‘From-Corpus-to-Cognition Principle’: “Frequency in text instantiates entrenchment in the cognitive system.”
Chapter 3 Aspects of description: ditransitive verbs in language use This chapter provides a corpus-based description of individual ditransitive verbs in actual language use. First, the two verbs that are typical of ditransitivity in ICE-GB will be analysed: give and tell (see section 3.1). Second, the four habitual ditransitive verbs in ICE-GB (i.e. ask, show, send and offer) will be scrutinised (see section 3.2). Particular emphasis in all the analyses will be placed on the different kinds of routines that are involved in the use of ditransitive verbs. The description of peripheral ditransitive verbs, on the other hand, will centre on the concepts of grammatical institutionalisation and conventionalisation (see section 3.3). At the end of this chapter, the two aspects will be discussed in a wider setting in the assessment of the role of linguistic routine and creativity in the use of ditransitive verbs (see section 3.4). 3.1
Typical ditransitive verbs in ICE-GB
In the present study, typical ditransitive verbs are verbs which are frequently attested in ICE-GB in general (i.e. > 700 occurrences) and which are associated with an explicit ditransitive syntax in some 50% of all occurrences or more (cf. Figure 2.4, p. 84). These standards are met by give (see section 3.1.1) and tell (see section 3.1.2). 3.1.1
GIVE
In light of recent psycholinguistic and cognitive-linguistic evidence, it is not surprising that the most frequent ditransitive verb in ICE-GB is GIVE.1 Experimental data have led Ninio (1999), for example, to put forward the hypothesis that children initially acquire constructions through one (or very few) ‘pathbreaking verbs(s)’. For the ditransitive construction, it stands to reason that GIVE, on grounds of its semantics, could be one of the pathbreaking verbs.2 Furthermore, the very prototypicality of GIVE may well be the reason why this verb is the most 1
From now on, the abstract lemma will be given in capital letters, while actual word-forms of the lemma will be given in lower case and in italics. 2 Note, however, that Campbell and Tomasello (2001) find that children may also start using the double-object construction with less prototypical verbs, which might be due to the input provided by their parents. The issue of prototypicality will be discussed in more detail in section 4.2.1.
Aspects of Description
93
frequent verb in the basic ditransitive pattern both in many children’s and in many adults’ language use (cf. Campbell and Tomasello, 2001: 258f.). As a matter of fact, GIVE is used not only in the basic ditransitive pattern, but in a variety of syntactic patterns. It is the range of these patterns and their frequencies in ICE-GB that will be discussed first (see section 3.1.1.1). In a second step, I will turn to significant principles and factors that are responsible for the selection of one specific pattern in given contexts (see section 3.1.1.2). 3.1.1.1 Structure and frequency of GIVE-patterns In the basic ditransitive pattern, all three semantic roles (i.e. acting entity, affected entity and transferred entity) are ‘explicitised’ at the surface-structure level (as subject, indirect object and direct object respectively).3 What is more, all clause elements are realised as noun phrases. Examples (20) to (22) are instantiations of the basic ditransitive pattern, which I refer to as the type-I pattern of GIVE.4 (20) (21) (22)
and then you can give everyone a runtime version of it On Tuesday members of Parliament gave the government their overwhelming support Moreover, Irish voters have wisely never given him an overall parliamentary majority
The subject may at times not be part of the type-I pattern, for example in imperatives and in occurrences of GIVE that are embedded in to-infinitives or ing-clauses. This is the case in examples (23) to (25). (23) (24) (25)
Give me a warning next time The theory behind these manoeuvres is to give other road users clear indication of your intentions By giving Patricia a grey skin and and the yellow and the green in the face etc you’re saying this is not about her real appearance
If we allow for the general optionality of the grammatical subject in such cases, the type-I pattern can be described by using the formula in (26).5 3
I am taking over the technical term ‘explicitise’ from computer linguistics where it is used to refer to the formal realisation of (i.e. ‘making explicit’) an underlying process or phenomenon or implicit relationship (cf. Lonsdale et al., 2001). 4 It should be noted that the distinction of patterns here is much more fine-grained than, say, the traditional ‘clause patterns’, because, for example, each clause pattern (being defined on the basis of configurations of clause elements) covers a wide range of different realisations of a particular clause element (cf. Biber et al., 1999: 141). 5 Note that ‘(...)’ refers to optional clause elements and ‘[...]’ to obligatory clause elements.
94 (26)
Chapter 3 I (S) GIVE [Oi:NP] [Od:NP]
From this basic type-I pattern, several other patterns can be derived, see (27) to (45) below. In type Ia, the direct object is in postverbal position, whereas the indirect object is in final position. Type Ib refers to a relative-clause structure in which the former direct object is fronted in order to serve as an antecedent to which the relative clause (with the remaining elements of the GIVE-pattern) refers back.6 Type Ic is similar, but it is the original indirect object which now serves as the antecedent for a relative clause. Note that in types Ib and Ic, the subject of the relative clause is an obligatory element. In type Id, the direct object is simply fronted. Type IP is the passive pattern in which the indirect object becomes the passivised subject. The direct object remains in situ, and the byagent may or may not be added. Finally, there is type IPb, which is based on the passive form IP. However, the direct object is now used as an antecedent to which a relative clause (with the remaining elements of the pattern, i.e. the passivised subject, the verb and the optional by-agent) refers back. In the following, the formulas and some illustrative examples are given for each of the six derivative patterns. What they all have in common with the basic type-I pattern is the explicitation of both the affected and the transferred entity as two objects and their realisation as noun phrases.7 (27) (28)
I a (S) GIVE [Od: NP] [Oi:NP] ‘He’s my dog. You gave him me.’
(29) (30) (31)
I b [Od: NPantecedent] (rel. pron.) [S] GIVE [Oi:NP] Those batteries that you gave me lasted an hour I nodded my head earnestly to wipe out any bad impressions I had given him previously
(32) (33) (34) (35)
6
I c [Oi:NPantecedent] (rel. pron.) [S] GIVE [Od: NP] ... to tell the difference between something that’s very white which you could give a gradience value of nought ... Uh an American lady that I gave a lecture uhm on architecture ...
In a similar vein to the common optionality of the grammatical subject in the type-I pattern, no distinction is made between relative clauses with a relative pronoun and zero relatives. Pattern-wise, the relative pronoun is considered to be a syntactically optional element, the presence or absence of which neither affects the order and realisation of elements in the pattern at hand nor the left-to-right flow of patterns in the text. 7 In elements such as [Od:NPantecedent] and [Oi:PPto], the subscript refers to a specific form or function of the clause element, e.g. an NP functioning as antecedent and a PP introduced by to respectively.
Aspects of Description
95
(36) (37)
I d [Od: NPfronted] [S] GIVE [Oi:NP] Somebody she gave me
(38) (39)
IP [S < Oi active] BE given [Od:NP] (by-agent) One was that I was being given the opportunity to uhm learn and develop uhm physical skills You ’ve been given the answers already ... and multi-national companies were given favourable conditions to invest in the countries ...
(40) (41) (42) (43) (44) (45)
IP b [Od:NPantecedent] (rel. pron.) [S < Oi] BE given (by-agent) I enjoyed the time that I was given to to study and to explore And uh this the letter that the UN Secretary General has been given uh by John McCarthy The hernia was due to uh vomiting as a result of the treatment that she was given for her injury ...
When placing the indirect object in final position, language users rarely use pattern Ia (see (28), which is the only example of the type-Ia pattern in ICE-GB). Rather, it is customary to introduce the indirect object in end-position by a preposition, most commonly to. This, then, is the type-II pattern: the direct object comes first and is realised as a noun phrase, the indirect object comes second and is realised as a prepositional phrase. The formula for this pattern and some examples are given in (46) and (47) to (49) respectively. (46) (47) (48) (49)
II (S) GIVE [Od:NP] [Oi:PPto] I meant to give it to you earlier Well perhaps you ’re giving too much to other people In September 1921, the British newspapers gave prominence to one story
As in the type-I patterns, there are a couple of derivative patterns that are systematically related to the type-II pattern in (46). In type IIa, the indirect object is realised by a prepositional phrase, introduced not by to but by for. 8 In type IIb, 8
Note that (52) and (53) can also be seen as examples of the pattern ‘V NP for NP to-inf.’ which resembles the pattern ‘V-link ADJ for NP to-inf.’ as in it was easier for you to go home (cf. Mair, 1987; Erdmann, 1997; Wagner, 2000). These examples illustrate the fact that in natural speech, we usually have a continuous ‘pattern flow’ (Hunston and Francis, 2000: 215; Hunston, 2002: 146), often resulting in an overlap of patterns (see section 3.1.1.2). It should also be noted that in examples (51) to (53) multiple analysis is possible. Specifically, the for-phrase may also be seen as a postmodification of the preceding noun phrase. Multiple analysis of this structure exerts no influence on the overall interpretation of the data because the type-IIa pattern is a marginal phenomenon (see Table 3-1 on p. 99).
96
Chapter 3
which is reminiscent of type Ib in (29), the original direct object is shifted into front position and functions as an antecedent to which a relative clause, including the subject, GIVE and the indirect object, refers back. In type IIc, the indirect object (still realised as a to-phrase) is found before the direct object. Types IIa, IIb and IIc are described and exemplified in the following. (50) (51) (52) (53) (54) (55) (56) (57) (58) (59) (60)
II a (S) GIVE [Od:NP] [Oi:PPfor] The defendant had been ordered to attend at Leeds Crown Court to give evidence for the prosecution in a fraud trial The government has given the go-ahead for shops and garages to charge higher prices to customers using credit cards ... outside City Hall where the Mayor has given permission for them to sleep II b [Od:NPantecedent] (rel. pron.) [S] GIVE [Oi:PPto] ... because every penny we give to the visual arts and the performing arts it goes through the Arts Council ... One of the great things about by-elections [...] is of course this extra bit of power that they give to a voter Do they see the value of this or are they slightly bemused by the emphasis you give to to this topic here II c (S) GIVE [Oi:PPto] [Od:NP] Zax, you angel: I fear events have conspired against me ... to give to you an impression, which would be ... It was the Queen of course who gave to Norman Schwarzkopf the knighthood that makes him now Sir Norman
From type II, the passive form IIP can be derived. Here, it is the direct object that becomes the subject of the passive. Again, a further passive pattern exists in ICE-GB, namely type IIPb, which combines the structure of type IIP with a relative clause or a past participle construction that refers back to an antecedent.9 (61) (62) (63) (64)
9
IIP [S < Od active] BE given [Oi:NPto] (by-agent) I accept that an explanation was given to the Liverpool Crown Court but ... that explanation was wholly insufficient more time should have been given to sanctions It is not clear that enough thought has been given to the consequences of these proposals ...
Note that in the formula of the type-IIPb pattern, the relative pronoun is referred to as ‘(S < Od)’. It is of course co-referential with the antecedent, as shown by the subscript ‘co’. However, in the past participle construction in (68), the passivised subject ‘(S < Od)’ and the form of BE do not appear, which are therefore marked as optional elements in the pattern formula in (65).
Aspects of Description (65) (66) (67) (68)
97
IIP b [antecedent]co (S < Od)co (BE) given [Oi:NPto] (by-agent) ... there are Home Office advices which are given to Chief Constables ... I’ll leave you with the final message that was given to the world leaders at the summit The evidence given to Cullen was monitored closely ...
While in all type-I and type-II patterns all semantic roles of GIVE are explicitised at the level of surface structure, type-III patterns are marked by the omission of the indirect object. In other words, the affected entity of the GIVING event is not made explicit. However, if ditransitivity is regarded as a more or less stable lexicosemantic property of the ditransitive verb, these instances of GIVE should also be taken as examples of ditransitivity. From different perspectives, it has been argued that such omissions of specific elements of the GIVING event may be due to the fact that they are recoverable from the context, inferrable from world-knowledge or that their specification is irrelevant in a given context (see for example Matthews’s (1981) notion of syntactically ‘latent objects’, Newman’s (1996) concept of cognitively ‘unfilled elaboration sites’ and Thompson and Hopper’s (2001) remarks on ‘object deletion’). The bottom-line (and the point of view to which I also subscribe) is that GIVE always triggers an event type at a cognitive level which includes three semantic roles – whether or not all semantic roles are explicitised.10 This is also captured by the working definition of ditransitive verbs in (11), according to which any verb that occurs in the basic form of ditransitive complementation (e.g. GIVE in the type-I pattern) is a ditransitive verb in all other complementation patterns (e.g. GIVE in type-III patterns) as well (cf. section 1.3.1). The basic type-III pattern is described and exemplified in (69) and (70) to (72) respectively.11 (69) (70) (71) (72)
10
III (S) GIVE [Od:NP] Oi he wanted physical love and I couldn’t give that So for instance we can give a very nice account of coarticulation ... Figures 3.2, 3.3, and 3.9 give examples of data produced by a geostationary platform
See also the discussion of Jackson (1990) and Biber et al. (1999) in sections 1.2.4 and 1.2.6 respectively. 11 The notation ‘Oi’symbolises the omission of the indirect object, although the corresponding semantic role of GIVE, i.e. the affected entity, is taken to be implicitly evoked at a cognitive level. Note again that the subject in the type-III pattern is marked as ‘optional’ in order to account for imperatives and other clause structures without grammatical subjects, see (23) to (25).
98
Chapter 3
From the basic type-III pattern in (69), some additional patterns can be derived. In type IIIb, the direct object serves as an antecedent for a relative clause (including the subject and GIVE). Type IIIP is the passive form in which the direct object becomes the subject of the passive. Type IIIPb is a pattern which combines passivisation with a relative clause or a past participle construction: the subjectivised direct object is now realised as a relative pronoun – or a zero-form – that is co-referential with an antecedent for a subsequent relative clause or past participle construction. The three derivative type-III patterns are described and exemplified in the following.12 (73) (74) (75) (76) (77) (78)
(79) (80) (81) (82) (83) (84)
III b [Od:NPantecedent] (rel. pron.) [S] GIVE Oi The sermon he gave when his daughter was married I suppose that statistic which you gave eighty per cent of the population wanting to go into the countryside ... Now in the long answer that he gave half way down that page he said towards the end ... IIIP [S < Od active] BE given Oi (by-agent) ... the only occasion on which the Reith Lectures have been given by a man of religion was in the late nineteen seventies when Dr Edward Norman took as his theme Christianity and the World Order Uhm here the treatment is given every week for twelve weeks More specific implementation details are given at the end of the report IIIP b [antecedent]co (S < Od)co (BE) given Oi (by-agent) But what I have simply done is to trace on to a map the directions that are given which give you some indication ... ... it also is of relevance when considering the evidence given by Mr Holt because there is a clear conflict ... and he will also know of the increased uh support given uh in the uh announcement last week by my right honourable friend the Social Security Secretary
If all occurrences of GIVE are to be regarded as instances of ditransitivity, consequently those cases have to be taken into consideration as well in which neither object is explicitised. This state of affairs is captured by type IV, as shown in (85) and exemplified in (86) to (88).
12
Note again that in the formula of the type-IIIPb pattern, ‘(S < Od)’ refers to the relative pronoun.
Aspects of Description (85) (86) (87) (88)
99
IV (S) GIVE Oi Od If you give and take when there’s that close bodily contact it’s great Builders always give The other major point he raises is in addressing the question of ‘why give in the first place?’
Table 3-1 provides an overview of the frequencies of all GIVE-patterns in ICE-GB. The two left-hand columns describe the individual patterns as presented above, the two right-hand columns give the number of occurrences and the relative percentages of each pattern in relation to the sum total of 1064 instances. Table 3-1: type I Ia Ib Ic Id
Frequency of GIVE-patterns in ICE-GB pattern
(S) GIVE [Oi:NP] [Od:NP] (S) GIVE [Od: NP] [Oi:NP] [Od: NPantecedent] (rel. pron.) [S] GIVE [Oi:NP] [Oi:NPantecedent] (rel. pron.) [S] GIVE [Od: NP] [Od: NPfronted] [S] GIVE [Oi:NP] miscellaneous IP [S < Oi active] BE given [Od:NP] (by-agent) IPb [Od:NPantecedent] (rel. pron.) [S < Oi] BE given (by-agent) II (S) GIVE [Od:NP] [Oi:PPto] IIa (S) GIVE [Od:NP] [Oi:PPfor] IIb [Od:NPantecedent] (rel. pron.) [S] GIVE [Oi:PPto] IIc (S) GIVE [Oi:PPto] [Od:NP] miscellaneous IIP [S < Od active] BE given [Oi:PPto] (by-agent) IIPb [antecedent]co (S < Od)co (BE) given [Oi:NPto] (by-agent) miscellaneous III (S) GIVE [Od:NP] Oi IIIb [Od:NPantecedent] (rel. pron.) [S] GIVE Oi miscellaneous IIIP [S < Od active] BE given Oi (by-agent) IIIPb [antecedent]co (S < Od)co (BE) given Oi (by-agent) IV (S) GIVE Oi Od miscellaneous sum
sum
%
404 1 23 2 1 10 84 12 123 4 7 2 6 23 17 2 247 16 3 38 28 10 1 1064
38.0 0.1 2.2 0.2 0.1 0.9 7.9 1.1 11.6 0.4 0.7 0.2 0.6 2.2 1.6 0.2 23.2 1.5 0.3 3.6 2.6 0.9 0.1 100
Note that phrasal verbs such as GIVE AWAY, GIVE IN and GIVE UP have not been taken into account in Table 3-1, because their semantics tends to be quite different from GIVE. Note also that not all occurrences of GIVE can be grouped
100
Chapter 3
into one of the patterns described above. In all pattern types (i.e. I to IV), there is a tiny remainder which is subsumed into the category ‘miscellaneous’.13 Since particular emphasis should be placed at this stage on more or less frequent patterns, these instances of GIVE do not pose any serious problems. What is more, of the 18 patterns described above, eight patterns alone make up for 91.2 % of all instances of GIVE in ICE-GB. These eight patterns are visualised in Figure 3-1: I, Ib, IP, II, IIP, III, IIIP and IIIPb. The entirety of the eight most frequent patterns that account for more than 90% of all occurrences of GIVE is related to a 2%-threshold level for each individual pattern since each of these eight patterns alone refers to at least 2% of all cases under scrutiny. 38.0%
23.2%
11.6% 8.7%
7.9%
Figure 3-1:
O
th er s
2.2%
I Ty pe IIP
2.2%
Ty pe
b
2.6%
III P
Ty pe
IP Ty pe III P
Ty
pe
II Ty
pe
II Ty pe I
Ty pe
I
3.6%
The eight most frequent GIVE-patterns
In the following, I will be concerned with the eight most frequent patterns of GIVE and the factors that are responsible for their lexicogrammatical routinisation. The underlying assumption, then, is that a model of lexicogrammatical routines in language use should not attempt to explain each and every occurrence in performance, but abstract away from the performance data a model that is able to account for some 90% of all cases. This approach capitalises on Mindt’s (2002) empirical redefinition of the notion of ‘grammatical rule’. He convincingly argues 13 This category includes, for example, sentences such as But almost as if she had some sort of message to give me (, ‘I misc.’), what are we going to give ourselves to start this song (, ‘II misc.’), …as if he didn’t know what to give… (
Aspects of Description
101
that even in the realm of syntax proper, a grammatical rule usually covers some 95% of all performance data, while some 5% of the instances may defy the rule on grounds of processes of ongoing language change, linguistic creativity, intended ungrammaticality and other factors involved. This remainder in performance that cannot be captured by a clear-cut rule is what he calls ‘the compost of the language’.14 Mindt’s (2002) dynamic understanding of grammar, which can be extended to the field of lexicogrammar, is in accordance with J. Aarts’s (1991) plea for models of language use that filter out from the entirety of performance data what is frequent and normal (see section 2.3). By definition, the compost of the language is neither frequent nor normal (although it may be at the basis of newly evolving rules and routines from a diachronic point of view). With regard to the eight most frequent GIVE-patterns at hand, there are also plausible functional reasons why it is these patterns that have become routinised in language use. These functional reasons are at the heart of the following section. 3.1.1.2 Principles of pattern selection for GIVE Table 3-1 (p. 99) shows that GIVE is not exclusively associated with one particular pattern (say, the basic complementation in type I), but with a variety of patterns. Even if we confine ourselves to the eight patterns that account for some 90% of all instances of GIVE, one principal question needs to be answered: which principles and factors cause language users to choose a specific pattern? As for the lexicogrammar of GIVE, the basic type-I pattern can be regarded as the default case both structurally and quantitatively (cf. Table 3-1, p. 99). It is thus of particular interest to describe functional reasons why language users opt for patterns other than this default pattern in specific contexts.15 For type Ib, one specific factor can be easily identified. This type tends to be used whenever the transferred entity of the GIVING event (which would translate into the direct object in the basic type-I pattern) has already been activated in the preceding text because it is part of the pattern that precedes GIVE. As shown in (89), this explanation accounts for some 83% of all cases. In examples (90) to (92), the previous pattern is italicised, and the GIVE-pattern at hand is underlined. These examples illustrate the fact that a preceding pattern in the text (e.g. the ... the ..., grateful for sth., thank so. for sth.) predetermines the following pattern by way of providing the initial slot and element for the next pattern. In effect, then, subsequent patterns are amalgamated with one another. In examples (90) to (92), type Ib is thus chosen because it makes it possible to start off from the transferred entity. 14 Consider in this context Halliday’s (1992: 65ff.; 1993) hypothesis that the 9:1 ratio is central to many aspects of language. 15 For GIVE and all other typical and habitual ditransitive verbs, the discussion of patterns will be structured as follows: (1) identification of the default pattern of the verb at hand; (2) discussion of all patterns that belong to the same type as the default pattern; (3) other types in order of their structural similarity to the default pattern.
102 (89)
Chapter 3 I b [Od: NPantecedent] (rel. pron.) [S] GIVE [Oi:NP] part of a previous pattern (19 of 23 cases = 82.6%)
(90) (91) (92)
But it then means that the more things they put on the menu the tinier the amount they give you I would anticipate doing one or two units per year and would be grateful for any financial assistance that the college could give me I must thank you, Simon and your parents ‘officially’ for the slow cooker and table cloth you gave us for our wedding
For the passive type IP, there are many factors that play a role in the process of pattern selection. The cluster of relevant factors is summarised in (93). What is not surprising is the fact that in more than 96% of all instances, the byagent is left out. A crucial factor in choosing type IP thus lies in the optionality of the agent.16 Additionally, two further factors seem to be responsible for the fact that the affected entity (corresponding to the indirect object in the default active type-I pattern) is placed in the initial slot, thus serving as the grammatical subject. First, this pattern tends to be chosen whenever the direct object is significantly heavier than the initial element and is therefore placed in final position according to the ‘principle of end-weight’ (Quirk et al., 1985: 1362). This correlation between weight and pattern selection is illustrated in examples (94) and (95). This factor seems to exert an influence on 50% of all 84 cases. Second, in some 10% of all cases it is the affected entity that has already been activated before and is thus taken up as the first element. In examples (95) to (97), the previously activated element which is part of (or provides the initial element for) the GIVEpattern at hand is italicised. (93)
IP [S < Oi active] BE given [Od:NP] (by-agent) activated before/ taken up (8 of 84 cases = 9.5%)
(94) (95)
16
heavy left out (42 of 84 (81 of 84 cases cases = 50.0%) = 96.4%)
... Margaret Thatcher cannot be given all the credit for our record levels of radioactivity both at sea and on land and rather nastily she had been tied to a chair until she was fourteen by her blind mother and never actually given any form of uhm sound or language communication
In many standard grammars, the omission of the by-agent has in fact been described as the unmarked state of affairs for the passive voice in general (cf. Quirk et al., 1985: 164f.).
Aspects of Description (96) (97)
103
After all Saddam Hussein uh led his people they although they were not given much choice in the matter in an eight year war ... The Italian peoples were bound to fight in Rome’s wars at their own charge ... Some peoples were actually given Roman citizienship ...
Next, I would like to turn to the type II-patterns. In type II again, it is a cluster of factors that can be shown to come into play to different extents in the process of pattern selection. As has been shown in (46), type II differs from the basic type I in that the indirect object is realised as a prepositional phrase, introduced by to, and placed after the direct object. Heaviness of the final element is again a relevant factor since it is involved in 39 of 123 cases (= 31.7%). But there is another factor that seems to be even more important for language users’ choice of this pattern in given contexts, namely the lexical item in direct-object position. The lexical items that are frequently used as direct objects in type II can be grouped into three major types. In nearly 25% of all cases, it is the pronoun it.17 The second group contains words which are habitually associated with the preposition to according to the pattern information in the corpus-based Macmillan English Dictionary (Rundell, 2002), e.g. access, answer and reaction. These words have a pattern themselves which could be described in COBUILD manner (cf. e.g. Sinclair, 1995) as ‘N to n’ (e.g. comfort to so.). The third group includes words that are so closely associated with this pattern that the resulting word-pattern combinations may be regarded as lexically stabilised idioms, e.g. give birth to so./sth. and give rise to so./sth.: here, the type-I pattern no longer provides a genuine alternative.18 In particular, the meaning of GIVE is metaphorically extended in these idiomatic phrases since the transferred entity (e.g. birth, rise) is no longer a physical object. These three groups of lexical items in direct-object position account for some 75% of all occurrences of this pattern. The two factors that are relevant to the preference of the type-II pattern over others – namely weight of the indirect object and lexis of the direct object – are summarised in (98). The examples given in (99) to (101) are intended to illustrate the latter factor in particular. The lexical items in direct-object position that seem to trigger the selection of the type-II pattern are italicised: in (99) it, in (100) the noun comfort, and in (101) the noun rise. Additionally, the relevant small-scale pattern (with a noun and the preposition to) is given in boxes in (100) and (101).
17 As a matter of fact, many of these instances could also be explained by considering weight alone since the to-phrase is mostly significantly heavier than simple it. Consider example (99). In a sense, then, no single principle of pattern selection should be seen as winning out over another factor. Rather, they often tend to come into operation along with each other. 18 Note that while give so. comfort is perfectly acceptable, *give so./sth. rise is not.
104 (98)
Chapter 3 II (S) GIVE [Od:NP] [Oi:PPto] heavy (39 of 123 cases = 31.7%) frequent lexical items in Od-position (91 of 123 cases = 73.9%): 1. it (30 of 123 cases = 24.4%) 2. words that are associated with the preposition to in general, e.g. access, aid, answer, attention, comfort, consideration, credence, (one’s) name, reaction, reply, substance (18 of 123 cases = 14.6%) 3. words bound to type II in lexically stabilised idioms, e.g. give birth / rise / thought / way to so./sth. (43 of 123 cases = 34.9%)
so we can have an acid and alcohol and give it to the esterase which is a useful product (100) A clutch of opinion polls gave comfort to both sides in the simmering civil war yesterday (101) but when you follow that through you’ve got the means to give rise to a change in the method of accounting that’s adopted in the company
(99)
As has been shown in (61), from the type-II pattern the passive type-IIP pattern can be derived, in which the indirect object is still realised as a to-phrase. Despite the fact that this pattern occurs fairly infrequently, a functional analysis of the 23 instances in ICE-GB highlights some factors that may be involved in the selection of the type-IIP pattern.19 In fact, it seems as though all the factors that are involved in the choice of the passive pattern IP are also involved in type IIP: previous activation of the initial element (6 of 23 cases = 26.1%), heaviness of the post-verbal element (8 of 23 cases = 34.8%), and the frequent omission of the by-agent (22 of 23 cases = 95.6 %). While the first two factors explain why either type IP or type IIP is selected (because the initial and postverbal elements are different), the last factor is the reason why in general a passive pattern is used in the first place. It is thus a reasonable assumption to take the relevance of these factors for the selection both of the type-IP and the type-IIP pattern for granted, although the database for the type-IIP pattern is comparatively thin. In the light of the 23 cases at hand, we may also assume that two further factors may at times tip 19 It is obvious that from a database of some 20 instances alone it is not possible to derive statistically significant trends. However, it will become obvious in the course of the present chapter that many factors that are being tentatively suggested for GIVE in this section are also relevant to the selection of equivalent patterns of other ditransitive verbs. Despite the necessity to restrict the manual corpus analysis to a 1-million-word corpus (see section 2.2.3), the functional analysis in the present chapter may still be seen as being able to offer an explanation as to why specific patterns are preferred over others in given contexts.
Aspects of Description
105
the balance in favour of type IIP: (a) the need to put the indirect object in focus according to the ‘principle of end-focus’ (Quirk et al., 1985: 1357); (b) the use of a lexical item (e.g. thought) in the passive subject which may be habitually associated with the preposition to. The cluster of all five factors and their explanatory power in quantitative terms are summarised in (102). Example (103) illustrates the relevance of the principle of end-focus (here in order to contrast the two italicised elements at the end of the two dependent clauses). Example (104) refers to the influence of the lexis in subject position on the selection of the type-IIP pattern. (102)
IIP [S < Od active] BE given [Oi:PPto] (by-agent) activated before/ taken up (6 of 23 cases = 26.1%)
heavy left out (8 of 23 (22 of 23 cases cases = 34.8%) = 95.6%)
words that are associated with the preposition to [cf. Macmillan English Dictionary] (9 of 23 cases = 39.1%)
deliberately placed in final focus position (7 of 23 cases = 30.4%)
(103) At the start of the conflict you said more time should have been given to sanctions but now you’re saying that more time should have been given to pursue those diplomatic initiatives (104) It is not clear that enough thought has been given to the consequences of these proposals for the movement of traffic outside the areas immediately affected As shown in Figure 3-1 (p. 100), the type-III pattern, in which the affected entity is not explicitised, is the second most frequent pattern of GIVE in ICE-GB. In roughly a quarter (23.2%) of all instances, the typical ditransitive verb GIVE is associated with a syntactically two-valent pattern. It is not surprising that corpus data reveal that this pattern is used whenever the affected entity is indeed recoverable from the context, inferrable from world-knowledge, or whenever its specification is irrelevant in a given context. Examples (105) to (107) illustrate such contexts in which there is no further need for the language user to explicitly state the affected entity of the GIVING event. Specifically, the affected entity has already been mentioned in the preceding text (see elements in italics). (105) he wanted physical love as well and I couldn’t give that (106) on economic adventures to the Soviet Union. Is it not the case that the British Government is far more reluctant to step in and give
106
Chapter 3
economic assistance than the countries of Germany, France and Italy (107) But it is a charity that needs your help and I hope you’ll give it In the light of linguistic research into the general differences in lexis and grammar between spoken and written language (cf. e.g. Chafe and Danielewicz, 1987; Halliday, 1987; Miller and Weinert, 1998; Cornbleet and Carter, 2001), one could easily hypothesise that the omission of the affected entity – and thus the use of the type-III pattern – is typical of spoken language, because the real-time constraints could lead language users to leave out particular elements that are recoverable from the shared situational context.20 However, Table 3-2 shows that for the verb GIVE in ICE-GB, the omission of the affected entity is even more frequent in writing than in speaking if the total number of occurrences of the type-III pattern in the two media are standardised to frequencies per million words. Table 3-2:
Frequency of the type-III pattern of GIVE in the spoken and written section of ICE-GB
type-III pattern of GIVE total number of occurrences frequency (per million words)
spoken section
written section
137
110
214.8
259.6
It transpires that not only the recoverability of the affected entity from the context (or, one should add, from world-knowledge) leads language users to choose the type-III pattern. Additionally, there is a tendency for language users to opt for type III whenever the lexical item in direct-object position, referring to the transferred entity, does not necessitate any specification of the affected entity, because it is only the mere existence of an affected entity that is relevant but not the particular kind of entity. The underlying reason may well be lexicosemantic in nature: in phrases such as give a lecture and give a message the items lecture and message always imply some kind of affected entity (i.e. an audience or a recipient of whatever kind). Since the existence of an affected entity as such is already implied in the situation type evoked by the ditransitive verb GIVE, the affected entity does not have to be explicitised at the syntactic surface structure (i.e. as an indirect object). In (108), 21 lexical items are listed that are used at least three times in ICE-GB as direct objects in the type-III pattern of GIVE. Note that these 21 words alone account for roughly 50% of all cases of this pattern. Some of these items that frequently occur in the type-III pattern are exemplified in (109) to 20
This hypothesis is also based on the assumption that typical spoken language is much more contextualised than writing in the sense that writing – unlike speaking – is usually permanent and distant.
Aspects of Description
107
(111). Generally speaking (and in a similar vein to the type-II pattern), specific words may thus serve as lexical pointers to specific patterns.21 (108)
III (S) GIVE [Od:NP] Oi contextually recoverable / specification irrelevant (all 247 cases = 100.0%) frequent lexical items (≥3): account (9), birth (3), command (3), detail (10), effect (3), evidence (20), example (9), hint (3), impression (10), indication (7), information (5), instruction (5), it (4), lecture (8), message (3), (so.’s) name (4), notice (3), reason (3), signal (3), talk (3), way (6) (124 of 247 cases = 50.2%)
(109) So for instance we can give a very nice account of coarticulation ... (110) It helps to clarify the poet’s ambiguous comments beforehand by giving an actual example of what he means (111) Figures 3.2, 3.3, and 3.9 give examples of data produced by a geostationary platform From type III, the passive form IIIP can be derived. Again, the optionality of the by-agent is of paramount importance in the process of pattern selection because it is omitted in 31 out of 38 cases (81.6%). Additionally, specific lexical items in the subject position (i.e. the subjectivised direct object of the type-III pattern) tend to be closely associated with this pattern. In other words, not only is the type-IIIP pattern used whenever neither the acting entity nor the affected entity needs to be explicitised; also, particular words referring to the transferred entity may serve as lexical pointers to this pattern. In (112) those words are listed that occur at least twice (i.e. recurrently) in this pattern in ICE-GB, accounting for some 45% of all instances. Some of them are exemplified in (113) to (115). (112)
IIIP [S < Od active] BE given Oi (by-agent) left out (31 of 38 cases = 81.6%) recurrent lexical items (≥2): approval (2), limit (2), information (2), detail (7), time (2), directions (2) (17 of 38 cases = 44.7%)
21
Many of these lexical items could be complemented with other items of the same semantic field that also occur in this GIVE-pattern in ICE-GB, e.g. give a lecture/a talk (+ a paper, a speech, a statement...), give instructions (+ advice, help, orientation...) and give a message (+ an answer, an outline, a response, a warning...). The important point here is that the lexis in direct-object position seems to be semantically restricted, which is reminiscent of Renouf and Sinclair’s (1991) concept of ‘collocational frameworks’.
108
Chapter 3
(113) He’s called Malachi in the opening verse but no biographical information is given about him (114) uh directions are given from Ushant uh from the Scillies uh from the South coast of Ireland down to Cape Ortegal or Finisterre (115) More specific implementation details are given at the end of the report The last pattern to be mentioned is type IIIPb. This type is similar to pattern Ib in that the transferred entity (i.e. the subjectivised direct object) serves as an antecedent to which a relative clause or a past participle construction refers back. As shown in (116), there is again a clear tendency for language users to choose this pattern with an antecedent (naturally placed in front position) whenever this antecedent has already been part of a preceding pattern in the text at hand. Examples (117) to (119) illustrate this dependency on the previous pattern (see patterns in italics: know of sth., consider sth., trace on to sth. sth.), the last element of which provides the starting-point for the subsequent GIVE-pattern. It should be noted in passing that the by-agent is not as frequently omitted as in all other passive patterns mentioned so far. In fact, in more than one third of all cases (10 of 28 cases = 35.7%), the agent is explicitly stated. Thus, the optionality of the by-agent as such turns out to be less forceful a factor for this particular passive form. (116)
IIIP b [antecedent]co (S < Od)co (BE) given Oi (by-agent) part of a previous pattern (16 of 28 cases = 57.1%)
with or without agent (10 vs. 18 cases = 35.7% vs. 64.3%)
(117) and he will also know of the increased uh support given uh in the uh announcement last week by my right honourable friend the Social Security Secretary (118) ... it also is of relevance when considering the evidence given by Mr Holt because there is a clear conflict ... (119) But what I have simply done is to trace on to a map the directions that are given which give you some indication ... From a functional point of view, it is not at all surprising that the eight GIVEpatterns described and exemplified above are the most frequent ones. In a sense, they represent eight different building blocks, all of which are associated with the verb GIVE. The main difference between the eight patterns can be systematised along the lines of two variables: (a) the syntactic explicitness of semantic roles and (b) the linear arrangement of semantic roles. According to the first variable, the eight GIVE-patterns can be subsumed into four different groups:
Aspects of Description (a) (b) (c) (d)
109
patterns in which all three semantic roles are explicitised (types I, Ib, II); patterns in which the acting entity tends to be omitted (passive types IP and IIP); patterns in which the affected entity is not stated explicitly (type III); patterns in which the acting entity is usually omitted and the affected entity is left out (passive types IIIP and IIIPb).
In other words, language users choose a specific pattern of one of the four groups according to the different needs in different contexts to explicitise the various semantic roles that are inherent in the GIVING event. The second variable can be used to distinguish between patterns on the grounds of the different arrangements of the explicitised semantic roles. In Table 3-3, the eight GIVE-patterns are systematised along the two variables. Table 3-3:
Explicitness and arrangement of semantic roles in the eight most frequent GIVE-patterns
GIVE-pattern type I type Ib type IP type II type IIP type III type IIIP type IIIPb
position 1
position 2
position 3
ACT DV AFF ACT DV ← TRA DV TRA ← AFF ACT DV TRA DV AFF ← TRA ACT DV TRA TRA DV (ACT) DV (ACT) ← TRA ACT – acting entity; AFF – affected entity; DV – ditransitive verb; TRA – transferred entity
position 4 TRA AFF (ACT) AFF (ACT) -------
For each pattern in Table 3-3, the explicitised semantic roles are indicated in the position in which they occur. Semantic roles in brackets indicate their optionality (and, in fact, their frequent omission). The arrow is intended to visualise the more or less frequent ‘anchoring’ of the affected/transferred entity in the preceding text, whenever it is shifted to front position.22 Such anchoring may be based on the previous activation of the initial element of the GIVE-pattern (usually in types IP and IIP) and/or on the fact that the initial element of the GIVE-pattern is also the last element of a preceding pattern (usually in types Ib and IIIPb). The overview in Table 3-3 thus reveals that each pattern is unique with regard to the 22
The notion of ‘anchor’ is also used by Prince (1981) – but also others, e.g. Birner and Ward (1998) – in order to refer to a specific kind of discourse familiarity of a textual element. Specifically, she distinguishes between so-called ‘brand-new’ elements and ‘newanchored’ elements, the latter being more familiar to the recipient because of their cognitive/semantic anchoring in the co-text or context.
110
Chapter 3
explicitness and/or arrangement of the semantic roles. The only exceptions are patterns IIIP and IIIPb, which do not differ as to those two variables. However, they are functionally different because in type IIIPb, unlike type IIIP, the initial element is anchored in the preceding text.23 Although, in general, explicitation and arrangement of semantic roles turn out to be very powerful principles of pattern selection for language users’ choice of a specific GIVE-pattern, the corpus analysis has shed light on several other factors that may come into play in the selection of individual or several patterns. These additional factors are of four different kinds: (1) the preceding pattern in the text, the last slot/element of which serves as the starting-point for the GIVEpattern which offers an appropriate arrangement of items; (2) specific words, which may be associated more or less closely with – and which thus function as ‘lexical pointers’ to – a specific pattern;24 (3) the principle of end-focus, which may lead to the choice of a GIVE-pattern that makes it possible to adhere to this principle; (4) the principle of end-weight, on the basis of which a GIVE-pattern may be preferred that places the most complex element at the end. The principal assumption, then, is that the eight GIVE-patterns under scrutiny compete with each other and that the above-mentioned principles of patterns make language users opt for a specific pattern in a given context: different GIVE-patterns are different building blocks that language users have at their disposal whenever they want to verbalise a GIVING event.25 The important point is that the selection of a particular building block is not at all of a random nature, but can be shown to be highly principle-guided. The influence on the pattern-selectional process with regard to GIVE is visualised in Figure 3-2. The circle in mid-position represents the process of pattern selection and includes those eight patterns that are most readily available to the language user whenever he/she wants to verbalise a GIVING event. The two horizontal arrows refer to the text as such (preceding and following the GIVING event) and capture a phenomenon that has already been described above by capitalising on Hunston and Francis’s (2000) concept of ‘pattern flow’:
23
In both patterns, the transferred entity usually represents given information; the important point here is that in type IIIP, the initial element is usually not anchored as clearly and explicitly as in type IIIPb. It should be noted in passing that one logically possible arrangement of the three semantic roles is not among the frequent three-valent GIVE-patterns, namely the pattern with the following order: affected entity – acting entity – (ditransitive verb) – transferred entity. The corresponding type-Ic pattern occurs only twice in ICE-GB. 24 Note that this factor is different from Wasow and Arnold’s (2003: 132ff.) concept of ‘lexical bias’ of a given verb towards a specific pattern. On the contrary, I would tentatively suggest that a detailed analysis of principles of pattern selection renders it unnecessary to assume a general lexical bias of a verb as such towards a specific pattern because these principles and factors explain why a verb is used in a specific pattern in a particular context. 25 As will be discussed in section 3.4, such patterns provide flexible building blocks which are subject to considerable variation in language use.
Aspects of Description
explicitation of semantic roles
preceding text preceding pattern
111
arrangement of semantic roles
lexical pointers
type I pattern type Ib type IP pattern type II selection type IIP type III type IIIP type IIIPb
focus/ weight
subsequent text subsequent pattern
verbalisation of a GIVING event Figure 3-2:
Routinised pattern selection: the case of GIVE
Pattern flow occurs whenever a word that occurs as part of the pattern of another word has a pattern of its own. ... Pattern flow is an extremely common phenomenon and can be found in all kinds of speech and writing. (Hunston and Francis, 2000: 211f.) The concept of pattern flow is reminiscent of Chafe’s (1979) ‘flow model’ and mirrors Sinclair’s (1993) and Brazil’s (1995) use of ‘prospection’. Apart from the fact that Hunston and Francis (2000) describe pattern flow as an all-pervading and genre-independent phenomenon, it is important to note that not only the selection of a specific GIVE-pattern may be based on a preceding pattern, but that the selected GIVE-pattern in turn may have repercussions on the pattern to the right. This is visualised by the dotted arrow on the right-hand side of Figure 3-2. Apart from the constraints that stem from previous and subsequent patterns in a given text, there are also other factors that clearly exert an influence on the process of pattern selection: explicitation of semantic roles, arrangement of semantic roles, lexical pointers and focus/weight considerations. The final aspect that needs to be tackled is the question of what the relationship is between the frequency and use of the patterns of a ditransitive verb on the one hand and the spoken/written-distinction and genre differences on the other. The following section is devoted to this particular question both in a wider theoretical setting and with regard to the GIVE-patterns in particular.
112
Chapter 3
3.1.1.3 Genre distinctions in a model of language use: the case of GIVE In corpus linguistics, especially Biber’s (1988) seminal research into text-typological classification on grounds of linguistic features has sparked off an increasing amount of corpus-based studies of language variation across genres.26 In general, these studies have provided empirical testimony of the fact that many aspects of language use (in the field of, say, collocations, phraseology and grammatical constructions) should be described genre-specifically, since different genres differ considerably with regard to such aspects. Thus, Kennedy (1998) is certainly right in pointing out that a general description of a language in its entirety needs to be complemented by a more fine-grained account of genre-specific particularities: The point is that while generalizations about a language as a whole are desirable and necessary, such generalizations can conceal systematic differences which exist between registers. (Kennedy, 1998: 198) However, it is a matter of dispute which position and place should be ascribed to register/genre variations in the description of language use. On the one hand, Biber et al. (1999), for example, argue in the Longman Grammar that no linguistic description can manage without taking into consideration register differences: In most cases, it is simply inaccurate or misleading to speak of a general pattern of use for English; instead, each register has distinctive patterns, associated with its particular communicative priorities and circumstances. (Biber et al., 1999: 24) On the other hand, Halliday (1992) takes a more balanced view in that he regards both a language-as-a-whole and a genre-centred description as viable – and not at all mutually exclusive – since the former approach provides a picture of the global norm of English while the latter approach may shed light on the extent to which specific genres may, for various reasons, deviate from the general norm. It 26
It is not my intention to discuss the problem of demarcation between the notions of ‘text-type’, ‘genre’ and ‘register’ here. Suffice it to say that I follow Biber and Finegan (1986) in distinguishing between genres as intuition-based and externally defined textgroups on the one hand and text-types as linguistically motivated and internally defined text-groups on the other. Note that in his later work “all aspects of variation in use” are subsumed under the notion of ‘register’ by Biber (1995: 9). Unfortunately, this terminology conflates the domains of genre and text-type to some extent. The terminological issue is further complicated by the fact that in the Longman Grammar, Biber et al. (1999: 15) use ‘register’ as a synonym for ‘genre’: “Register distinctions are defined in nonlinguistic terms”.
Aspects of Description
113
seems to me that, in the final analysis, Halliday’s (1992) line of argumentation is that language is not the sum of its registers, but that a specific register is a particular kind of language. As a matter of fact, the question of whether language is the primary notion and register the secondary one or vice versa involves a chicken-or-egg style predicament. Due to the unanswerability of this very question, it is possible and reasonable to describe language use at both levels, i.e. globally and genre-specifically. Halliday (1992) describes the misunderstanding on which the exclusive focus on genres is based as follows: There has been some misunderstanding on this topic, with the argument being put forward that since every text is in some register or other only register-based frequencies have any meaning, and it is meaningless to talk of global probabilities in grammar. This is rather like saying that since it is always spring, summer, autumn or winter a mean annual precipitation is a meaningless concept. (Halliday, 1992: 68) I agree with Halliday (1992) on the point that it useful to establish a norm for the language as a whole on the grounds of ‘global probabilities’ obtained from general and representative corpora in their entirety. Language is a system of linguistic forms and structures that is based on an abstraction of actual usage. A genre may best be regarded as a contextually and situationally defined subsystem of language which also represents an abstraction from usage and which is characterised by a specific patterning of the forms and structures that the language system provides. In other words, the structural possibilities that are available in English lexicogrammar are rooted in the language system and not in a particular genre. As far as lexicogrammar is concerned (unlike, say, the level of textlinguistics and pragmatics), genres are clearly secondary to the language system. Of course, even at the level of lexicogrammar a descriptive focus on the language system does not rule out more specific research into genre-specific trends. But I would contend that such genre-specific trends should not be at the core of a lexicogrammatical model of language use, as it has been sketched out for the verb GIVE and its complementation in Figure 3.1 (p. 111). More specifically, any discernible genre-specific deviation from the global probabilities in the use of GIVE (as described in section 3.1.1.2) should be regarded as being based on a contextual strengthening or weakening of the principles of pattern selection that are already part of that model of language use: (1) explicitation of semantic roles; (2) arrangement of semantic roles; (3) preceding pattern; (4) lexical pointers; (5) focus/weight. In specific genres, some factors turn out to be more important, while other factors are more relevant to other genres. In the following, I will sketch out how the model of language use suggested in Figure 3.1 (p. 111) could be easily complemented by a genre-specific description (including the more general spoken/written-distinction), which, however, does not call into question the plausibility of the general and genre-independent model as such.
114
Chapter 3
The first question that needs to be answered in this context is of course the following one: is there a correlation between the frequency of GIVE-patterns and the distinction of spoken and written language (which may be taken to represent the perhaps most general genre distinction)? Table 3-4 gives the results of a chisquare (χ2) test.27 While the observed frequencies represent the actual numbers of occurrences of the patterns at hand in ICE-GB, the expected frequencies are calculated for each cell by multiplying the row total and the column total and dividing it by the grand total of 970 instances (cf. Butler, 1985: 118ff.). For each pair of observed and expected frequency, the formula (O-E)2/E is then used to compare the deviation of observed frequencies from the expected values. The sum of all (O-E)2/E values is the test statistic χ2, which in Table 3-4 is 28.77. Table 3-4:
GIVE-patterns in the spoken and written section of ICE-GB: a chisquare test
GIVEpattern observed
spoken expected
written (O-E)2/E
observed
I 258 232.8 146 2.73 Ib 17 13.3 6 1.03 IP 42 48.4 42 0.85 II 73 70.9 50 0.06 IIP 9 13.3 14 1.39 III 137 142.3 110 0.20 IIIP 13 21.9 25 3.62 IIIPb 10 16.1 18 2.31 column (876.5 559 411 total per mill. words) chi-square 28.77 degrees of freedom 7 significance level (0.05 level) ≥ 14.07
expected
(O-E)2/E
171.2 3.71 9.7 1.41 35.6 1.15 52.1 0.08 9.7 1.91 104.7 0.27 16.1 4.92 11.9 3.13 (970.0 per mill. words)
row total 404 23 84 123 23 247 38 28 970
In Table 3-4, the number of degrees of freedom involved is (8-1) × (2-1) = 7 since [it] is determined by the fact that if we know all but one values in a row or column, and also the row or column total, the remaining cell entry can be worked out. Therefore degrees of freedom = (no. of rows – 1) × (no. of columns – 1). (Butler, 1985: 121) 27 As a matter of fact, other statistical measurements of distribution and significance such as the t-test (cf. Butler, 1985: 84ff.) would also be possible. However, the point that I want to make in this section, i.e. that genre-specific comparisons between actual language data and abstract norm distributions warrant an in-depth linguistic interpretation and may hide general facts about the underlying language system, is independent of any individual statistical test.
Aspects of Description
115
Given the number of degrees of freedom, the significance level (at a five percent level) is taken to be 14.07 (cf. Butler, 1985: 176). Since χ2 in Table 3-4 is 28.77, the observed distribution and the expected distribution of the GIVE-patterns in the spoken and in the written section are taken to be significantly different; thus, the null hypothesis can be safely rejected. Note, however, that χ2 is a cumulative value to which specific values of (O-E)2/E contribute much more than others. For example, the effect of (O-E)2/E for the type-IIIP pattern in the written section (4.92) on χ2 is much greater than that of (O-E)2/E for the type-II pattern in the spoken section (0.06). On the basis of the observed and expected frequencies for the eight most frequent GIVE-patterns, the conclusion has to be drawn that there is a statistically significant association between the spoken/written-distinction and the frequency of the GIVE-patterns (at the five percent level). However, I do not think that this association invalidates the language-as-a-whole model of the lexicogrammar of GIVE as described in section 3.1.1.2. Rather, in order to account for the spoken/ written-variation in the use of the GIVE-patterns, it is reasonable to start off from the very model suggested in Figure 3-2 (p. 111) and to refine it by analysing the different extents to which specific principles of pattern selection are relevant to spoken and written language. Specifically, the difference in distribution of GIVEpatterns between spoken language and written language is a trend that stems from different extents to which system-internal factors come into play in the two media. That there is a difference between the observed and the expected frequency of each individual GIVE-pattern, as shown in Table 3-4, does not come as a surprise. As Hunston (2002: 161) points out, “variation is found wherever it is sought”, and this is particularly true of a comparison between actual data and statistical norm distributions. Since χ2 is a cumulative test statistic, it is important to note that a particular group of GIVE-patterns contributes very much to the overall fact that there is a statistically significant correlation between the distribution of the GIVE-patterns on the one hand and the spoken/written-distinction on the other. Specifically, it is the difference between observed and expected frequencies in the passive patterns that is largely responsible for the χ2 value of 28.77. In other words, if all passive patterns were more or less as frequently used as statistically expected, the value of (O-E)2/E would be 0.00 for all passive patterns. In this hypothetical case, the resulting χ2 would be entirely based on the active patterns and their values of (O-E)2/E, which would be 9.49 (2.73 + 1.03 + 0.06 + 0.20 + 3.71 + 1.41 + 0.08 + 0.27) and thus clearly below the significance level of 14.07. This reading of Table 3-4 shows that it may well be the active/ passive-distinction that is primarily responsible for the association between the distribution of the GIVE-patterns on the one hand and the spoken/writtendistinction on the other. In Table 3-5 below, the result of a second chi-square test is given in which the focus is on the correlation between the frequency of active GIVE-patterns (types I, Ib, II, III) vs. passive GIVE-patterns (types IP, IIP, IIIP, IIIPb) on the
116
Chapter 3
one hand and the spoken/written-distinction on the other. It vindicates the conclusion that this correlation is a major factor indeed: χ2 is 19.03, which is far beyond the significance level of 3.84 (at the five percent level). Table 3-5: GIVEpattern
The correlation between the active/passive and spoken/written factors: a chi-square test spoken observed
expected
active 485 459.3 patterns passive 74 99.7 patterns column 559 total chi-square degrees of freedom significance level (0.05 level)
written (O-E) /E
observed
expected
(O-E) /E
row total
1.44
312
337.7
1.96
797
6.62
99
73.3
9.01
173
2
411
2
970
19.03 1 ≥ 3.84
The question arises if – and how – the two statistically significant correlations reported on in Tables 3-4 and 3-5 are related to each other. In my view, it makes sense to assume that the influence that the active/passive-distinction exerts on the different GIVE-pattern distribution in spoken and written language is hidden in Table 3-4. Table 3-5 sheds some quantitative light on this hidden factor. The important point here is that the influence of the active/passive alternation on the use of the GIVE-patterns has already been implicitly covered by the principles of pattern selection that are taken into account in the general model of the lexicogrammar of GIVE as visualised in Figure 3-2 (p. 111). In particular, I am thinking here of the two factors of explicitation of semantic roles and arrangement of semantic roles. Since in most passive patterns the acting entity is almost always left out, a passive pattern is used in given contexts because it allows for the omission of the optional by-agent. This of course is related to the selection of a pattern in which the arrangement of semantic roles is such that the first element is either the transferred entity or the affected entity (cf. Table 3-3, p. 109). I would thus contend that it is indisputably useful to analyse in detail the distribution of GIVE-patterns in spoken and written language but that the factors responsible for quantitative differences are medium-dependent specifications of the general and medium-independent principles of pattern selection that are relevant to the lexicogrammar of GIVE in general.28 Broadly, a language-as-a-whole model of lexicogrammatical patterns (and the factors involved in the process of pattern selection) is not invalidated by the fact that differences between spoken and written language do exist. But these differences are neither categorical nor an 28
For a detailed discussion of the notions of ‘medium-dependent’ and ‘medium-independent’, see Esser (2000b).
Aspects of Description
117
integral part of the language system but are based on different extents to which system-internal patterns and principles of pattern selection are utilised in the two media. A similar problem is the question whether genre distinctions should be at the basis of any model of language use or whether it is possible to develop a general and genre-independent model (say, of the lexicogrammar of GIVE) first which could then be complemented with more detailed analyses of specific genres. Again, I think that the general principles of pattern selection as suggested in section 3.1.1.2 have a great explanatory power when it comes to the routinised use of GIVE-patterns in the English language in its entirety. It would again be possible to delve more closely into genre-dependent particularities. However, as I see it, such genre-specific features would very often turn out to be caused by factors that are specifications of the general principles of pattern selection as suggested in Figure 3-2 (p. 111). One case in point is the use of the pronoun it as direct object in the type-II pattern, i.e. ‘GIVE it to...’. It was suggested in (98) that the pronoun it as direct object is a significant factor that triggers off the selection of the type-II pattern of GIVE. A syntactic reason is of course, as indicated in (98), that the indirect object is very often heavier than it and is therefore placed in final position, realised as a to-phrase, as for example in (99).29 Additionally, one could easily hypothesise that ‘GIVE it to...’ prevails in spoken genres, because the communicative situation is much more likely to provide information on what the pronoun it is intended to refer to. Table 3-6 shows that in fact more than 75% of all instances of ‘GIVE it to...’ occur in the spoken component of ICE-GB. Interestingly, some 70% of all spoken instances can be found in one genre alone, namely in direct conversations. Table 3-6:
Distribution of ‘GIVE it to...’ in ICE-GB: some quantitative trends
‘GIVE it to...’ ⇒ ICE-GB ⇒ spoken component ⇒ direct conversations (S1A)
occurrences 30 23 16
frequency (100 %) ⇒ 76.7 % ⇒ 69.6 %
It stands to reason that for all GIVE-patterns such genre-specific findings can be described on the basis of corpus data. Do such findings invalidate a general model of the lexicogrammar of GIVE, based on language-as-a-whole principles of pattern selection? I would say no. Note that in Figure 3-2 (p. 111), one general principle of pattern selection is circumscribed as ‘lexical pointers’. In this con29
Even if the indirect object is not heavy (and is not a pronoun either), it in direct-object position triggers off the type-II pattern: “If Od is a personal pronoun, the prepositional construction is favoured, especially if the other NP is not a pronoun – examples like %I gave Kim it are inadmissible for most speakers, especially in AmE” (Huddleston and Pullum, 2002: 310). Thus, the fact that the pronoun it is a lexical pointer to the type-II pattern seems to stem from a variety of reasons, including not only heaviness but also focus constraints.
118
Chapter 3
text, the pronoun it is a lexical pointer to the type-II pattern. Why it is used for the transferred entity in the first place, is another question altogether and has to do with communicative and situational factors that may vary considerably from genre to genre. I would contend that such genre-specific factors could be easily added to the periphery of the model in Figure 3-2 (say, by including a more specific subcomponent within ‘lexical pointers’), but they do not shake the viability of the genre-independent core of the model. The growing interest in spoken/written and genre variation in language use is no doubt a welcome development, given the fact that many aspects of this dimension of language variation have gone unnoticed so far. However, language is more than the sum of its genres, and the core of the underlying language system should, in my view, be regarded as a genre-independent abstraction which is bound to be realised in the spoken or written medium and in one genre or another. This implies that there certainly is no straightforward answer to the question whether genres are the product of language use or whether language use is the product of genre distinctions. It seems to me that a corpus-based quantitative and functional analysis of language use and the deduction of a general model of language use is perfectly possible without focusing on differences between spoken and written language or between specific genres. I have already argued elsewhere that verb complementation is a good case of largely genreindependent and verb-specific choices and preferences (cf. Mukherjee, forthcoming), which is also corroborated by Biber et al. (1999: 388): “variation between verbs is far greater than any differences across registers.” Thus, a model of verb complementation in general and ditransitive verb complementation in particular may well be aimed at the system of language as a whole. It is this system which provides the linguistic patterns and basic principles of pattern selection that may turn out to be more or less important for either medium and/or any particular genre. As pointed out by Esser (1993: 170f.), the corpus-based description of the language system results in a ‘heterogeneous corpus norm’ for the language as a whole, while the corpus-based analysis of a specific genre provides a more ‘homogeneous corpus norm’ for the genre at hand. The two kinds of abstracted corpus norm are not mutually exclusive but account for different aspects of quantitatively feasible trends in language use, namely global norms and local norms respectively. At least at the level of lexicogrammar, there are many principles at work that are genre-independent so that the appropriate starting-point for a model of language use should be a broad language-as-a-whole perspective rather than a narrow genre-specific analysis. Besides, the issue of genre distinctions (e.g. genre definitions in general and the delicacy of genre distinctions in particular) will always be a matter of dispute (cf. Hunston, 2002: 161). Thus, two different levels of language (and norm) description should be kept apart: (1) global norms of language in its entirety (‘description1’); (2) local norms in terms of, say, spoken, written and genre-specific trends (‘description2’). It is the global norms in the routinised lexicogrammar of ditransitive verbs (in terms of description1) that are also at the heart of the following section, in which I will turn to the verb TELL.
Aspects of Description
119
3.1.2 TELL Among others, Halliday (1994) links ditransitive verbs to three different (yet cognitively similar) process types: ‘material processes’, ‘verbal processes’ and ‘relational processes’ (see section 1.2.4). While GIVE, in a sense, represents the prototypical example of material processes (i.e. situation types) with three participants (i.e. semantic roles) involved, TELL may be regarded as prototypical of three-role verbal processes since it is the second most frequent ditransitive verb in ICE-GB. In this context, Newman (1996: 136ff.) argues that there is a connection between the situation types underlying GIVE and TELL in that the semantics of TELL as a verbal-communication verb is similar to a specific metaphorical extension of (literal) GIVE. This can be observed in a “schematic network associated with a GIVE predicate in those languages where the predicate is also used in the sense of TELL (with appropriate objects)” (Newman, 1996: 138). For English, this metaphorical extension from literal GIVE to a verbal-communication GIVE and its resemblance to verbal-communication TELL can be shown by referring to the description of the situation types in (120) to (122): (120) literal GIVE: Æ acting entity TRANSFER transferred entity (physical object) TO affected entity e.g. in He gave her the keys. (121) metaphorical extension of GIVE to verbal communication: Æ acting entity TRANSFER transferred entity (verbal message) TO affected entity e.g. in He gave her a lecture. (122) literal TELL: Æ acting entity TRANSFER transferred entity (verbal message) TO affected entity e.g. in He told her a story. The kinds of transfer processes involved in GIVE and TELL thus shade into each other and are actualisations of the same abstract situation type with analogous semantic roles: “Telling something to someone thus amounts to the verbal equivalent of giving” (Newman 1996: 138).30 As was done for GIVE, I will first provide an overview of the TELLpatterns and their frequencies as attested in ICE-GB (see section 3.1.2.1). In attempting to develop a model of the lexicogrammatical routines involved in the use of TELL, I will then generalise from the corpus data significant principles of pattern selection (cf. section 3.1.2.2). 30
The cognitive implications of this GIVE-TELL-gradient (which indicates the prototypicality of the category of ditransitivity) will be discussed further in chapter 4.
120
Chapter 3
3.1.2.1 Structure and frequency of TELL-patterns The structurally – though not quantitatively – most basic pattern of TELL is the type-I pattern with both objects realised as noun phrases.31 The pattern formula is given in (123), and three illustrative examples can be found in (124) to (126).32 (123) I (S) TELL [Oi:NP] [Od:NP] (124) Later his assassin said he told us nothing (125) She greeted me with every sign of pleasure, and invited me to tell her the latest news from Oxford (126) Oh I tell you what For TELL, there are four variations of the type-I pattern that can frequently be found in ICE-GB. The TELL-pattern Ib is equivalent to the GIVE-pattern Ib, as represented and exemplified in (27) and (28) on p. 94. Here, the original direct object is shifted to front position and serves as an antecedent to which a relative clause (with the remaining elements of the TELL-pattern) refers back. Not surprisingly, we also find type-I patterns in which the direct object, representing the transferred entity (i.e. usually a verbal message), is realised as a wh-clause (type Ie), a that-clause (type If) or a to-infinitive (type Ig).33 The formulas for the four derivative type I-patterns and some examples are given in (127) to (142).34 (127) Ib [Od: NPantecedent] (rel. pron.) [S] TELL [Oi:NP] (128) but uh it was scored at a very reasonable rate which no doubt Ian will tell me in a moment (129) Having said that there are one or two things I want to tell you 31 Note that the pattern distinctions follow those introduced for GIVE in section 3.1.1.1. In particular, a TELL-pattern that is structurally identical with a specific GIVE-pattern is given the same coding (e.g. type I). 32 Example (126) shows that those instances in which the direct object is realised as what (or as so) are also subsumed under type I since from a syntactic point of view they could be replaced with, say, this or something. 33 Note that the introductory that in object that-clauses may be omitted (cf. Quirk et al., 1985: 1049). 34 From a traditional structuralist point of view, type Ie in fact falls into two types, i.e. finite wh-clause objects and wh-infinitive clause objects (cf. Quirk et al., 1985: 1214f.). On the other hand, Trotta (2000), in his very detailed study of wh-clauses, distinguishes between four subtypes (wh-interrogatives, wh-exclamatives, wh-free relatives and whbound relatives). For the purpose of the present study, however, the general category of wh-clauses seems appropriate. The broad distinction between that-clauses, wh-clauses and to-infinitives picks up on the categorisation of complement clauses that can be found in the Longman Grammar (cf. Biber et al., 1999: 660ff.). Note also that, syntactically, I regard the to-infinitive in type Ig as a direct object of TELL and, semantically, as part of the situation type of TELLING (in the role of the transferred entity).
Aspects of Description
121
(130) I mean I find it ridiculous the things they ’ve told me (131) Ie (S) TELL [Oi:NP] [Od:wh-clause] (132) Can you tell us how you first got involved in this project (133) After ten years of Mrs Thatcher, the government now openly censors the BBC and tells our teachers what they can and what they cannot teach our children (134) But you must at some time have told Mr Sainsbury who your solicitor was (135) If (S) TELL [Oi:NP] [Od:that-clause] (136) Tell him he needs it (137) When you go for your sight test, tell the optician that you think you can have a voucher (138) I accept that he was telling you that it was clearly to do with drugs (139) Ig (S) TELL [Oi:NP] [Od:to-infinitive] (140) Don’t tell me to keep talking or I’m going to keep quiet (141) I shall have to tell him not to try so much (142) After he married, he told his wife not to open the metal box because it belonged to a friend Within type I, various passive patterns can be distinguished, depending on the active type to which they correspond systematically. In type IP, the indirect object becomes the subject of the passive, see (143) to (146). In type IPa, the indirect object becomes the passive subject as well, but the direct object is now in fronted position, see (147) to (150).35 In types IPe, IPf, and IPg, it is again the indirect object which becomes the subject of the passive, but here the direct object (which remains in situ) is realised as a wh-clause, a that-clause and a to-infinitive respectively (in analogy with the active patterns Ie, If and Ig). Formulas and examples of these passive patterns are given in (151) to (161). (143) IP [S < Oi active] BE told [Od:NP] (by-agent) (144) Yes we were told that (145) And President Bush who’s spending the weekend at Camp David was woken up to be told the news of the attack on Tel Aviv 35
Note that for TELL (unlike GIVE) fronted what is also included in type IPa because the corresponding unmarked word order is in principle possible and acceptable for TELL (e.g. I tell you what) but not for GIVE (e.g. ?*I give you what).
122
Chapter 3
(146) therefore, although the gist remained it is likely that when recounting the deeds they were told the context of the period (147) IP a [Od:NP] [S < Oi active] BE told (by-agent) (148) and that I was told by the head of the department (149) No that’s what I was told (150) It is based on what I was told by several HODs ... (151) IP e [S < Oi active] BE told [Od:wh-clause] (by-agent) (152) You don’t want to be told how it’s scored (153) They were called media response teams which I thought was an interesting new development in which journalists were really carried around and shown what to report and told what to report effectively (154) IP f [S < Oi active] BE told [Od:that-clause] (by-agent) (155) So I ’ve been told I’ve got to do a month o o a month of reading (156) I ’m told that the general practitioner is now dead (157) The jury was told the event went off well and there was no trouble until it was winding up shortly before 3 am (158) IP g [S < Oi active] BE told [Od:to-infinitive] (by-agent) (159) We ’ve all been told to do it (160) so girls are told to wear dresses wear clothes and so forth (161) he was told to use the normal exit and that caused ‘resentment and friction’ For TELL it is also possible to place the indirect object in final position and realise it as a to-phrase. This is the type-II pattern. From this pattern the passive type IIP can be derived, in which the direct object becomes the subject and the tophrase remains in situ. While the two patterns are fairly frequent with GIVE (11.6% and 2.2% respectively, cf. Table 3-1, p. 99), both of them are attested with TELL only once in ICE-GB, see (162) to (165). (162) II (S) TELL [Od:NP] [Oi:PPto] (163) In order to test this, they set up an experiment within which two groups of undergraduates told stories to the rest of the class ... (164) IIP [S < Od active] BE told [Oi:PPto] (by-agent) (165) ... the story I have been telling you about the curious personal habits of our new Dean was told to the Bursar, in the strictest confidence, only this morning, and may well not be common knowledge until the middle of next week
Aspects of Description
123
Especially the low frequency of type II calls into question the widely held view that alternation between type I and type II is as common with TELL as it is with, say, GIVE. For example, Campbell and Tomasello (2001: 259), in their study of early acquisition of ditransitive verbs and their constructions, are surprised that “reasonably alternating verbs such as tell, feed, hand, and pay were limited to the double-object construction [i.e. type I] by many children.” As for TELL, however, this finding does not come as too much of a surprise since in adult language – and, we may assume, in caretaker language – type II is hardly ever used. In the light of the striking differences between GIVE and TELL with regard to the typeI/type-II alternation, it is also unfortunate that in many reference grammars both verbs are subsumed into the same complementation group (cf. e.g. Quirk et al., 1985: 1210; Huddleston and Pullum, 2002: 309).36 Type-III patterns are also significantly less frequently attested with TELL than with GIVE. In all type-III patterns, the indirect object is not explicitised. In the basic type-III pattern, the direct object is realised as a noun phrase and is in the unmarked postverbal position, see (166) to (169). In type IIIa, the direct object is realised as a wh-clause, see (170) to (173); it is realised as a that-clause in type IIIc, see (177) to (179). Type IIIb is equivalent to the type-IIIb pattern of GIVE in that the direct object is fronted and functions as an antecedent for a relative clause including TELL and its subject, see (174) to (176). (166) III (S) TELL [Od:NP] Oi (167) You can’t tell lies just because you have particular views (168) Yes it is completely unreal cos you know that next to nobody is actually telling the truth (169) While Isobel Lennox sleeps, downstairs in the Villa Cellini the Goldmans are telling jokes again partly in order to relieve tension (170) III a (S) TELL [Od:wh-clause] Oi (171) ... it’s quite hard to tell you know what is her being deliberately monosyllabic and what is her being downright miserable (172) We can’t tell offhand what’s interacting with what (173) and therefore they couldn’t tell really how much damage they had done 36
Note, however, that there is a difference in categorisation in Quirk et al. (1985: 1209f.) since they draw a structural analogy not only between type I and type II (pertaining to both GIVE and TELL), but also between tell to (i.e. type II) and tell about. However, the latter analogy is problematic from a semantic point of view as will be discussed later in this section. It should also be mentioned that Huddleston and Pullum (2002: 310) do acknowledge the fact that TELL – unlike most other verbs of verbal communication – is usually not used in the type-II pattern.
124
Chapter 3
(174) III b [Od:NPantecedent] (rel. pron.) [S] TELL Oi (175) Better than Snodgrass and the Marble Lamp which I was going to tell (176) and if they could get through to his characters and the story that he wished to tell he didn’t really mind at all what adults felt about them (177) III c (S) TELL [Od:that-clause] Oi (178) and I can tell by the look in his eyes that he didn’t actually but he then makes a mental note to do something about it (179) I can tell you’re waking up there As for passive patterns, TELL is associated with the same types as GIVE: type IIIP and type IIIPb as shown and exemplified in (180) to (187). (180) IIIP [S < Od active] BE told Oi (by-agent) (181) Fine dry and sunny we ’re told for the day (182) ... but the story is beautifully told in what is now considered to be the finest example of medieval stone carving in the world (183) ... because those types of stories are entertaining, and, when told properly, contain all the elements of good stories (184) IIIP b [antecedent]co (S < Od)co (BE) told Oi (by-agent) (185) ... this is one of the stories told by those princesses who were going to be set up in the palace as one saw in the other picture (186) A story which could no doubt be told by many children (187) Their hypothesis was that the construction of truth within the courtroom was primarily a matter of the overall narrative plausibility of the story told In type IV, both the direct and the indirect object are omitted. This pattern is shown in (188) and exemplified in (189) to (191). (188) IV (S) TELL Oi Od (189) But you could you could tell (190) Racing experience here will tell (191) We met a charming young man who was trying to write on the back of as far as I could tell an envelope one or two amendments to the Union constitution It has been noted in section 3.1.1.1 that GIVE very often occurs with the direct object only, especially in the type-III pattern, but that it is not attested in ICE-GB with the indirect object alone (cf. Table 3-1, p. 99). TELL, on the other hand, is fairly frequently attested with a pattern in which the indirect object (referring to
Aspects of Description
125
the affected entity) is explicitised but in which the direct object (representing the transferred entity) is not. This is the type-V pattern. From this pattern, it is also possible to derive a passive pattern VP: here, the indirect object becomes the subject. The type-V pattern and its passive counterpart VP are shown and exemplified in (192) to (199). (192) V (S) TELL [Oi:NP] Od (193) I ’ll tell you tomorrow (194) And she came and told her husband you see (195) Do we tell Simons? (196) VP [S < Oi active] BE told Od (by-agent) (197) and this way I can usually discover proposed future programmes all long before I ’d officially be told (198) I admit that I expected to be told oh last Friday evening or something of that sort (199) Her grandfather, the first Alistair MacIntosh, had been, she was told as a child, kidnapped and shipped to Argentine In categorising the patterns of TELL that can be found in ICE-GB in either of the types described and exemplified above, some problems and unclear cases do occur. While marginal unclear cases are subsumed into the category ‘miscellaneous’, a major problem remains when it comes to TELL in combination with direct/reported speech because the verbal message in direct/reported speech can be interpreted either as the direct object of TELL (resulting in, say, pattern If) or as a syntactically unrelated structure (resulting in pattern V). With TELL in combination with direct/reported speech, three linear configurations are in principle possible: a)
A tells B: “...” => Rajiv perhaps best captured the imagination of members of Congress when he told them: “India is an old country but a young nation …”
b)
“...”, A tells B => “An act of bravado,” she’d told Dr Rainbow
c)
“...”, A tells B , “...” => The late Mr Barnsley (Mr Pitkin had told her) was what is termed a self-made man
Although other syntactic analyses are no doubt possible, I have decided – as indicated by the boxes above – to consistently categorise the TELL-pattern in the three cases a) to c) as follows:
126 a) b)
c)
Chapter 3 If TELL introduces direct/reported speech, the direct/reported speech section has been regarded as the direct object of TELL, so that these instances of TELL are grouped into the pattern If or Ie. If TELL follows direct/reported speech, the direct/reported speech section has been considered as syntactically unrelated to the TELL-pattern (and not as, say, a fronted direct object of TELL). From a pattern-grammar perspective, it is assumed here that the TELL-pattern represents a selfcontained verbal process that is added to the verbal process inherent in the direct/reported speech section. In these cases the TELL-pattern is thus considered as an add-on to the direct/reported speech and is analysed as type V. If TELL interrupts direct/reported speech, the TELL-pattern is also analysed as referring to a self-contained verbal process and, accordingly, as a type-V pattern.
Phrasal and prepositional verbs involving TELL (e.g. TELL ABOUT, TELL FROM, TELL OF) have not been taken into consideration for the present study. As in the case of GIVE, such multi-word verbs have a different semantics from TELL proper. The most frequent case is TELL ABOUT. In the Comprehensive Grammar, TELL ABOUT is considered a second prepositional variant of TELL, with the about-phrase functioning as a prepositional object like the to-phrase in the type-II pattern (cf. Quirk et al., 1985: 1209f.). The example that the Comprehensive Grammar draws on in order to illustrate the structural analogy is the following one: tell [D1 + 2a + 2b]
Mary told only John the secret. [D1] Mary told the secret only to John. [D2a] Mary told only John about the secret. [D2b] (Quirk et al., 1985: 1209)
While the complementation patterns [D1] and [D2a] correspond to the type-I pattern and the type-II pattern respectively, I would contend that from a semantic point of view type [D2b] should not be regarded as a variation of [D1]. The important point here is that [D1] can be readily transferred to [D2a] and vice versa without changing the propositional content. In contrast, there is a clear propositional difference between telling someone a secret and telling someone about a secret (which may be paraphrased as informing about rather than literally telling someone a secret). Also, if secret is replaced with, say, friend, even the structural analogy suggested in the Comprehensive Grammar can no longer be drawn: tell someone about my friend is not transferrable to *tell someone my friend. For these reasons, I do not consider TELL ABOUT to be a variant pattern of the ditransitive verb TELL but as a multi-word verb in its own right. The same holds true for all other phrasal and prepositional verbs involving TELL. In Table 3-7 below, the frequencies of all TELL-patterns in ICE-GB are given. Disregarding the miscellaneous cases, there are nine patterns that make up
Aspects of Description
127
90% of all instances of TELL, cf. Figure 3-3. It is these nine patterns that section 3.1.2.2, which is devoted to the principles of pattern selection for TELL, will focus on. Table 3-7: type I Ib Ie If Ig IP IPa IPe IPf IPg II IIP III IIIa IIIb IIIc IIIP IIIPb IV V VP sum
Frequency of TELL-patterns in ICE-GB pattern
(S) TELL [Oi:NP] [Od:NP] [Od: NPantecedent] (rel. pron.) [S] GIVE [Oi:NP] (S) TELL [Oi:NP] [Od:wh-clause] (S) TELL [Oi:NP] [Od:that-clause] (S) TELL [Oi:NP] [Od:to-infinitive] miscellaneous [S < Oi active] BE told [Od:NP] (by-agent) [Od:NP] [S < Oi active] BE told (by-agent) [S < Oi active] BE told [Od:wh-clause] (by-agent) [S < Oi active] BE told [Od:that-clause] (by-agent) [S < Oi active] BE told [Od:to-infinitive] (by-agent) (S) TELL [Od:NP] [Oi:PPto] [S < Od active] BE told [Oi:PPto] (by-agent) (S) TELL [Od:NP] Oi (S) TELL [Od:wh-clause] Oi [Od:NPantecedent] (rel. pron.) [S] TELL Oi (S) TELL [Od:that-clause] Oi miscellaneous [S < Od active] BE told Oi (by-agent) [antecedent]co (S < Od)co (BE) told Oi (by-agent) (S) TELL Oi Od (S) TELL [Oi:NP] Od [S < Oi active] BE told Od (by-agent)
sum 96 8 97 215 26 19 6 3 2 37 5 1 1 27 10 2 2 4 4 5 11 116 11 708
% 13.6 1.1 13.7 30.4 3.7 2.7 0.8 0.4 0.3 5.2 0.7 0.1 0.1 3.8 1.4 0.3 0.3 0.6 0.6 0.7 1.6 16.4 1.6 100
Before turning to relevant principles of pattern selection for TELL, it is appropriate to compare the frequencies in Table 3-7 and Figure 3-3 with the quantitative data in the corpus-based Longman Grammar (cf. Biber et al., 1999: 388) in order to assess the overall reliability of the data obtained from ICE-GB. As pointed out in section 1.2.5, the Longman Grammar distinguishes between four genres: ‘academic prose’ (ACAD), ‘news writing’ (NEWS), ‘fiction writing’ (FICT) and ‘conversation’ (CONV). For each of these genres, the percentage of various complementation patterns of TELL is indicated. Note, however, that exact frequencies are not given. Rather, a distinction is made between six frequency ranges: (1) 75%, (2) 50-75%, (3) 25-50%, (4) 10-25%, (5) less than 10%, (6) not attested. Secondly, there is no one-to-one correspondence between
128
Chapter 3
30.4%
16.4% 13.7% 13.6% 10.0%
Figure 3-3:
1.6% 1.6%
I Ty pe IP f Ty pe III Ty pe Ig Ty pe IV Ty pe V P O th er s
3.8% 3.7%
Ty pe
Ie
V
Ty pe
Ty pe
Ty pe
If
5.2%
The nine most frequent TELL-patterns
the TELL-patterns in Table 3-7/Figure 3-3 and the complementation types that are distinguished in the Longman Grammar. Specifically, the Longman Grammar groups all clausal realisations of clause elements into the generic category ‘complement clause’. Also, no distinction is made between active patterns and their passive counterparts. In general, the complementation types in the Longman Grammar are defined much more broadly than in the present study. Notwithstanding these caveats, the quantitative data can be compared to some extent with the ICE-GB data as given in Table 3-7 and Figure 3-3. In Table 3-8, Biber et al.’s (1999: 388) findings are replicated. The corresponding TELL-patterns that are distinguished in the present study are given in the second row. Note that no correspondences are given for passive patterns because the Longman Grammar provides no information as to how the passive patterns have been categorised. Ignoring the passive patterns of TELL, however, does not distort the overall quantitative picture since they only account for some 10% of all instances of the verb TELL in ICE-GB (74 of 708 cases). Thus, a comparison of the frequency data in the Longman Grammar and in ICE-GB remains useful.37
37
Unfortunately, the Longman Grammar does not provide any quantitative information on the various verb complementation patterns of GIVE. It is for this reason that no such comparison between the Longman Grammar and ICE-GB could be offered in section 3.1.1.1.
Aspects of Description Table 3-8: Pattern in Biber et al. Pattern in Table 3-7 CONV FICT NEWS ACAD
129
Frequency of TELL-patterns in the Longman Grammar of Spoken and Written English (cf. Biber et al., 1999: 388)
SV
SVOd (NP)
SV + complement clause
type IV
types III, IIIb
types types IIIa, IIIc I, Ib
<10% <10% <10% <10%
<10% 10-25% <10% 25-50%
<10% not att. not att. not att.
SVOiOd (NP)
25-50% 25-50% <10% 10-25%
SVOdOi (PrepP)
SVOi + complement clause
SVOi
type II
types Ie, If, Ig
type V
25-50% 25-50% 75% 50-75%
25-50% 10-25% <10% <10%
not att. not att. not att. <10%
It is obvious that Table 3-8 sheds light on some important genre differences in the use of TELL-patterns. However, as suggested in section 3.1.1.3, there are also genre-independent trends in lexicogrammar. It is these language-as-a-whole trends that the present study is particularly interested in. Thus, a comparison between the quantitative data for TELL in the Longman Grammar and in ICE-GB should centre on two questions for the purpose of the present study: (1) Are the genre-independent trends in ICE-GB (as outlined in Table 3-7) comparable to the findings in the Longman Grammar in general? (2) What are these genre-independent trends in particular? The complementation type ‘SVOi + complement clause’, which is described as the by far most frequent one in the Longman Grammar, corresponds to the type-Ie, type-If and type-Ig patterns in Table 3-7. In ICE-GB, these three patterns alone account for 47.8% of all instances of TELL. The prevalence of these patterns in language use is thus corroborated by the Longman Grammar. Also, the very low frequency of the type-II pattern in ICE-GB is in line with the findings presented in the Longman Grammar: in CONV, FICT and NEWS, the corresponding complementation type ‘SVOdOi (PrepP)’ is not attested, and in ACAD it also found only sporadically. Broadly speaking, the similarly middleranking frequencies of the type-I pattern (13.6%) and the type-V pattern (16.4%) in ICE-GB, which correspond to the complementation types ‘SVOiOd(NP)’ and ‘SVOi’ respectively, are in line with the quantitative data provided by the Longman Grammar. Specifically, the Longman Grammar vindicates the general trend that type I and type V are similarly frequent, but are significantly less frequently used than the entirety of types Ie, If and Ig. The very low frequencies of the typeIV pattern (i.e. ‘SV’) and the type-IIIa and type-IIIc patterns (i.e. ‘SV + complement clause’) in ICE-GB are also in accordance with the findings in the Longman Grammar. The only striking difference between ICE-GB and the Longman Grammar refers to the type-III pattern, which corresponds to ‘SVOd (NP)’ in the Longman Grammar. While this type shows a comparatively low frequency in
130
Chapter 3
ICE-GB (3.8%), the Longman Grammar indicates at least for FICT and ACAD higher frequency ranges, namely 10-25% and 25-50% respectively. Apart from this difference, however, the general quantitative trends that are found in ICE-GB for TELL are corroborated by the quantitative data that are provided by the Longman Grammar. These trends can be summarised as follows: a) b) c) d)
TELL is very frequently used in type-Ie/If/Ig patterns; TELL is fairly frequently found in type-I and type-V patterns; TELL is rarely used in type-IV and type-IIIa/c patterns; TELL is hardly ever found in the type-II pattern.
These are some of the major, genre-independent trends in the lexicogrammar of TELL which should be covered by a viable model of language use. To this end, the following section is intended to unveil important principles and factors that are responsible for the selection of one particular TELL-pattern in a given context and, in a wider setting, for the quantitative trends as described above in general. 3.1.2.2 Principles of pattern selection for TELL The by far most frequent pattern of TELL is the type-If pattern, which accounts for 30.4% of all occurrences of TELL. As shown in (135) to (138), the direct object is realised as a that-clause in this pattern. It is quite clear that, from a strictly structural point of view, this pattern cannot be regarded as the unmarked choice, since type I – in which both objects are realised as noun phrases – is analytically simpler. On the other hand, from a quantitative perspective, type If clearly outnumbers type I (13.6%). Additionally, it is reasonable to regard type If as the default TELL-pattern because of the basic discourse function of the that-clause it includes: That complement clauses occurring in post-predicate position are commonly used to report the speech, thoughts, attitudes, or emotions of humans. In these constructions, the subject of the main clause refers to the human participant, the lexical verb or adjectival predicate presents the type of reporting (e.g. speech or thought), and the thatclause presents the reported speech, thought, or attitude ... . (Biber et al., 1999: 660) From such a discourse perspective, it does not come as a surprise that TELL tends to be used in type If, because TELL as a verb of verbal communication evokes a situation type in which the acting entity conveys a verbal message (i.e. the transferred entity) to the affected entity, see (122). The most typical way of verbalising this specific ‘verbal’ kind of transferred entity is a that-clause. Thus, the combination of frequency and discourse function of that-clauses is an important reason why the type-If pattern should be considered the lexicogrammatical default case
Aspects of Description
131
for TELL.38 The question arises when and why language users opt for other TELL-patterns. As shown in Figure 3-3 (p. 128), there are eight other TELLpatterns that can be found more or less frequently in ICE-GB, thus representing more or less routinised alternative patterns that are available to the language user. Three of the other routinised TELL-patterns do not differ from type If with regard to the arrangement and the explicitation of semantic roles: type I (13.6%), type Ie (13.7%) and type Ig (3.7%).39 The only difference between types I, Ie, If and Ig refers to different realisations of the direct object, representing the transferred entity: noun phrase, wh-clause, that-clause and to-infinitive respectively. Generally speaking, the four different realisations of the direct object correspond to four different discourse functions and, accordingly, different types of the semantic role that is fulfilled by the transferred entity. For the default pattern If, the specific kind of transferred entity can be described as a verbal message of a factual kind (MESSAGE/FACT).40 Note that this functional interpretation picks up on Dixon’s (1991: 237) definition of the meaning of the that-clause: “THAT complements refer to some assertable activity or state as a single unit, without any reference to its inherent constitution or time duration.” Language users thus tend to choose type If whenever the transferred entity is a fact in the sense of a piece of information (i.e. a state or activity) they want to get across as the acting entity’s verbal message. See examples (200) to (202):41 (200) You’d say I’m sorry but I have to tell you you’ve got a big globule of wax in your ear (MESSAGE/FACT) (201) Would you like to take a seat while I tell him you’re here (MESSAGE/ FACT) (202) She told me that she wants to be free of lawyers and doctors (MESSAGE/ FACT) ...
38 While that-clauses “refer to some definite event or state” (Dixon, 1991: 215), wh-clauses are used to inquire about some event or state. The discourse function of that-clauses and wh-clauses is thus similar, but since that-clauses (30.4%) are used much more frequently than wh-clauses (13.7%), it is useful to regard the type-If pattern as the default pattern of TELL. 39 Note that this is a fundamental difference between GIVE (cf. Table 3-1, p. 99) and TELL (cf. Table 3-7, p. 127). While the four most frequent patterns of GIVE – types I, IP, II and III – differ from each other with regard to the arrangement and/or the explicitness of semantic roles, these two factors are irrelevant at least to the four patterns of TELL at hand because all semantic roles are explicitised and arranged in the same order in the TELLpatterns I, Ie, If and Ig. 40 Quirk et al. (1985: 1212) describe this particular function of the that-clause as “introducing an indirect statement”. The second function they ascribe to so-called ‘that-clause objects’, i.e. as an indirect directive, will be taken up later in this section. 41 Note again that the type-If pattern includes that-clauses with and without the realisation of the conjunction that (cf. Quirk et al., 1985: 1049f.). However, the omission of that is a matter of stylistic variation and does not affect the selection of the type-If pattern – with its that-clausal realisation of the direct object – as such.
132
Chapter 3
In type Ie, on the other hand, the realisation of the direct object as a wh-clause is usually indicative of a different kind of verbal message, namely an answer to a real or hypothetical question (MESSAGE/ANSWER), or – in Dixon’s (1991: 237) words – “some aspect of an assertable activity or state ... about which clarification is needed.” Consider examples (203) to (205) in which the wh-clause could be easily rephrased as a question (say, by inversion) which would evoke the very answer that the TELL-pattern provides: what is it called? (203), why don’t you believe me? (204), who was Peel? (205). (203) I couldn’t tell you what it’s called (MESSAGE/ANSWER) (204) I can tell you why you don’t believe me (MESSAGE/ANSWER) (205) Now I’m sure I do not have to tell you who Peel was (MESSAGE/ ANSWER) In type Ig, the to-infinitive is used in direct-object position because the verbal message is neither a fact/statement nor a real/hypothetical answer but an order (MESSAGE/ORDER).42 More precisely, this meaning of the to-infinitive is described by Dixon (1991: 237) in terms of “the subject of the complement clause becoming involved in the state or activity referred to by that clause.” In other words, the recipient (i.e. the indirect object) is told to do something. Note that it is possible to paraphrase type Ig by using type If with a modal verb or a verb in the subjunctive mood in the that-clause.43 Consider examples (206) to (208). (206) ... because they told them to gather round water holes (MESSAGE/ORDER) => ... that they should gather round water holes (207) Her husband told her not to attend (MESSAGE/ORDER) and as a result the trial was impeded => ... that she should not attend ... (208) With one hand they’re telling you to reduce radiation at all costs (MESSAGE/ORDER) => ... that radiation be reduced at all costs In type I, which occurs 96 times in ICE-GB, the verbal message is not realised as a clause but as a more or less complex noun phrase with a central head noun or a pro-form. The verbal message is not presented as a process in the type-I pattern 42
Quirk et al. (1985: 1215) describe this function of to-infinitive clause objects as ‘indirect directive’. 43 It is due to the systematic correspondence between I told John to sleep (to-infinitive) and I told John that he should sleep (that-clause with modal verb) that Dixon (1991:45) calls this particular type of to-infinitive a ‘Modal (FOR) TO clause’. It is, of course, a matter of dispute whether there are compelling reasons for such an analogy between the type-Ig pattern of TELL and the for/to-pattern of other verbs (as in I wish for John to go).
Aspects of Description
133
but, by way of nominalisation, as a product (MESSAGE/PRODUCT): nouns and pro-forms that can be frequently found in direct-object position of type I include, for example, this/that (27), anything/something/everything/nothing (11), what (9), all (6), more (5), story (5) and truth (5). This list and the following examples highlight the fact that language users tend to choose the type-I pattern whenever they want to summarise the verbal message by nominalising it (e.g. story, truth) and, what is more, by very often using a context-dependent pro-form (e.g. anaphoric or cataphoric this, what).44 (209) I’ll tell you a funny story about working class (MESSAGE/PRODUCT) later (210) Mr Walling was not telling him the truth (MESSAGE/PRODUCT) (211) but tell me this (MESSAGE/PRODUCT) will they be about religion and do you expect them to be controversial (212) I tell you what (MESSAGE/PRODUCT) go and ask Toni where it is It is quite clear that especially two kinds of verbal messages – i.e. FACT and ANSWER – can be presented as a product by realising the indirect object as a noun phrase with an appropriate head. Examples (213a) and (214a) give examples of type-I patterns which could be regarded as nominalised presentations of the corresponding forms in (213b) and (214b). (213a) If by my merely telling you two facts (MESSAGE/PRODUCT) that Johnny went to Sam’s party and Sam blew out the candles ... (213b) ... telling you that Johnny went to Sam’s party and Sam blew out the candles (MESSAGE/FACT) (214a) ... and their predominantly martial character may tell us something about the ethos of Edward’s household in the 1290s (MESSAGE/PRODUCT) (214b) ... may tell us what the ethos of Edward’s household in the 1920s was (MESSAGE/ANSWER) What emerges from the categorisation above is thus a semantic principle that is responsible for the selection of any of the TELL-patterns I, Ie, If and Ig. The decisive factor here is the kind of transferred entity, which leads to the preference of a specific pattern over others. The four kinds of verbal messages that need to be distinguished (and that correspond to different TELL-patterns) are graphically visualised in Figure 3-4. 44
The correlation between context-dependence and spoken language will be discussed at the end of this section.
134
Chapter 3 transferred entity => MESSAGE
MESSAGE/ PRODUCT
MESSAGE/ FACT
MESSAGE/ ANSWER
MESSAGE/ ORDER
Æ [Od:NP] Type-I pattern
Æ [Od:that-cl.] Type-If pattern
Æ [Od:wh-cl.] Type-Ie pattern
Æ [Od:to-inf.] Type-Ig pattern
*Type-Ie pattern including [Od:wh-cl.] with exclamative function
*Type-If pattern including [Od:that-cl.] with modal verb/ subjunctive form *Type-Ie pattern including [Od:wh-cl.] with to-infinitive
Figure 3-4:
Correlations between TELL-patterns and specific kinds of verbal messages as transferred entities
Although the type-I pattern is the simplest – and most ‘basic’ – pattern from a merely structural point of view (i.e. with the direct object realised as a noun phrase just like the indirect object), the thick line in Figure 3-4 – connecting the transferred entity to the type-If pattern – indicates that from a quantitative and semantic point of view the transferred entity (which represents a verbal message in case of TELL) is most typically realised as a that-clause. That is, although the structurally basic pattern is type I, speakers tend to verbalise the transferred entity of the TELLING event as a that-clausal direct object by choosing the type-If pattern. Figure 3-4 shows not only the two patterns that are structurally basic (i.e. type I) and functionally basic (i.e. type If), but also the other two patterns in which the transferred entity corresponds to a specific kind of verbal message: in the type-Ie pattern, the verbal message is an ANSWER, and in the type-Ig pattern, it is an ORDER. Note that some cases of a clausal realisation of the direct object of TELL defy this clear-cut one-to-one mapping of a particular kind of transferred entity onto a specific pattern. Some patterns (i.e. types Ie and If) may also fulfil other functions under specific conditions (see the patterns marked with an asterisk in Figure 3-4). In particular, a that-clause with a modal verb or a subjunctive form is subsumed formally into that-clauses (and is thus taken to exemplify type If), but is semantically similar to a to-infinitive, see (215). Also, specific wh-clauses in type Ie do not indicate a MESSAGE/ANSWER but fulfil a different discourse function. As shown in (216), wh-clauses introduced by how or what may fulfil an exclamative function and present some sort of emotive statement, resembling to
Aspects of Description
135
some extent a MESSAGE/FACT (cf. Trotta, 2000: 101ff.). Also, as shown in (217), wh-infinitives present indirect directives and are thus functionally similar to a MESSAGE/ORDER.45 (215) Mr Ridley told the Commons they should pen more letters (≈ MESSAGE/ ORDER) => Mr Ridley told the Commons to pen more letters (216) I told them how much Thomas had enjoyed Trumpets in Grumpet Land (≈ MESSAGE/FACT) ... => I told them that Thomas had very much enjoyed Trumpets in Grumpet Land (217) Certainly moral philosophers could no longer tell us what to do (≈ MESSAGE/ORDER) => Certainly moral philosophers could no longer tell us to do anything These examples, however, are of a marginal nature and do not jeopardise the viability of the general form-function correspondences as visualised in the upper part of Figure 3-4. For example, wh-infinitives of the type exemplified in (217) account for only 8 cases of all 97 instances of the corresponding type-Ie pattern. Also, wh-clauses with an exclamative function are generally very rare in language use, as pointed out by Ellegård (1978: 29) and Trotta (2000: 101). Thus, even in the light of these special cases of the type-Ie pattern there is a strong correlation between the semantic role of MESSAGE/ANSWER and the formal realisation as a wh-clause in general. In essence, Figure 3-4 illustrates that the specific kind of transferred entity seems to be the most important reason why language users choose one of the patterns I, Ie, If or Ig in a particular context. The importance of this semantic principle of pattern selection is also corroborated by the following examples, in which two different TELL-patterns are used in exactly the same context. In (218), two different kinds of messages are presented (ORDER > to-infinitive and FACT > that-clause). In (219), one and the same message is at first presented as an answer to an anticipated question (ANSWER > wh-clause), but then as a factual – and more explicit – statement (FACT > that-clause). (218) The Interior Ministry has told people to carry on with their work (MESSAGE1/ ORDER) and that attempts to destabilise the country will be severely punished (MESSAGE2/FACT) (219) Dichter had told the Luccas what Mrs Lennox had said (MESSAGE1/ ANSWER) – that Isaac Simons had attacked her (MESSAGE1/FACT) 45
Dixon (1991: 237) describes the semantically intermediate position of wh-infinitives as follows: “WH-TO complements effectively combine the meanings of WH- and Modal (FOR) TO, referring to some activity in which the subject will get involved, and about some aspect of which clarification is needed.”
136
Chapter 3
These examples show that the particular semantics of the transferred entity clearly impinges on the TELL-pattern that language users tend to choose. While the kind of transferred entity is primarily responsible for the selection of a type-I pattern (i.e. I, Ie, If or Ig) in general, each of these – and all other frequent – patterns can also be described in more detail with regard to their individual features of linguistic routinisation. In the following overview, I will again regard the type-If pattern, which is the most frequent TELL-pattern, as the default case and will pay particular attention to lexical, semantic and pragmatic factors that are relevant to the choice of TELL-patterns other than the type-If pattern. Broadly speaking, the type-I pattern turns out to be highly routinised in terms of the lexical items that can be found in subject, indirect-object and directobject position. Specifically, in one third of all cases the subject is first person I (32 of 96 cases = 33.3%). In the indirect-object position, personal pronouns in general account for more than 90 percent of all instances of type I (88 of 96 cases = 91.7%).46 This tendency to use context-dependent pro-forms with the type-I pattern can also be found in the direct-object position since the most frequent NP heads here are the following ones: this/that (27), anything/something/everything/ nothing (11), what (9), all (6), more (5), story (5), truth (5), thing (4), a little bit (4).47 These words/expressions alone account for nearly 80 percent of all cases at hand (76 of 96 cases = 79.2%). Apart from the tendency of a small list of lexical items to be used as direct object in type I, it is particularly noteworthy that a substantial part of the meaning of these items relies on the given context and/or has to be made explicit somewhere else in the text, see examples (209) to (212) above. The frequently occurring nouns in direct-object position are always lexical items that represent the verbal message as a PRODUCT (e.g. story, truth); the verbal message itself (e.g. the actual story or the actual truth) is explicitised somewhere else in the text. Very often, pro-forms are used (e.g. anything, a little bit) that also signal to the recipient that the verbal message is made explicit elsewhere. In sum, then, the direct-object position in the type-I pattern is semantically restricted to nouns and to pro-forms that refer to verbal messages and whose meaning has to be inferred from the context. The same holds true for the prevalence of personal pronouns as indirect objects and the comparatively frequent use 46
The prevalence of personal pronouns as indirect objects and subjects in this pattern – and in many other TELL-patterns – is indicative of the fact that the corresponding semantic roles are usually animated. Since TELL is a verb of verbal communication – and verbal communication is a human trait – the animacy of the affected entity and the acting entity is not at all surprising. 47 It should be noted, however, that while this and that as direct objects are clearly referential items, story and truth as nominalised verbal messages are more salient semantically. With such nouns, it is also possible to expand – and thus make more explicit – the direct object by, say, adding a postmodifying prepositional phrase to the NP head, as for example in (209): a funny story about working class. But even in this largely context-independent indirect object, there remains a difference between the story about someone/something and the story itself (which would typically be introduced by a that-clause, cf. Figure 3-4, p. 134). In this case, context-independence should thus be taken to mean ‘independence of the original verbal message’.
Aspects of Description
137
of the first person pronoun I as subject in type I. The three restrictions on the lexical realisation of the three semantic roles of the type-I pattern are summarised in (220). (220)
I (S) TELL [Oi:NP] [Od:NP] frequent lexical items in Od-position (76 of 96 cases = 79.2%): this/that (27), anything/something/everything/ nothing (11), what (9), all (6), more (5), story (5), truth (5), thing (4), little bit (4)
I (32 of 96 cases = 33.3%)
personal pronouns (88 of 96 cases = 91.7%): you (39), me (24), us (11), her (7), him (5), them (2) From what is frequently attested in the various slots of the type-I pattern, I would suggest that a ‘pattern frame’ of conventionalised linguistic routines for this particular pattern can be abstracted away. Such a pattern frame, which is based on the summarised findings given in (220), is shown in Figure 3-5. The horizontal axis refers to the slots of the type-I pattern from left to right and the vertical axis to the frequency of specific lexical items in each slot. you me ... I
TELL
us her him ...
Figure 3-5:
this/that any-/some-/every-/nothing what all* more* story* truth* thing* a little bit* ...
... *about
A pattern frame for the type-I pattern of TELL
Figure 3-5 shows that, on the one hand, the type-I pattern can be regarded as a syntactic scheme that can, in principle, be used with an infinite set of lexical items in any of its slots (thus, ‘...’ can be found in all the slots). On the other hand, while there is a relatively great freedom of lexical choice in the subject position, there is much more lexical routine involved in the realisation of the indirect-object and the direct-object slot. Accordingly, ‘...’ is given on top of the subject slot because in two-thirds of all cases the lexical item in this slot is not routinised. On the other hand, ‘…’ is found towards the bottom of the directobject and indirect-object slots because it is only in comparatively few cases that the lexical items in these positions are not members of the lists given in (220). In
138
Chapter 3
the light of the lexical routinisation of the direct and the indirect object, it seems as though the type-I pattern provides a frame for a fairly restricted number of conventionalised lexicalisations of its syntactic slots.48 The pattern frame in Figure 3-5 thus illustrates the ‘unrandomness’ in the distribution of words (Sinclair, 1991: 110) for the type-I pattern of TELL. However, this type of partially fixed, conventionalised lexicalisations of the pattern at hand needs to be distinguished from clear-cut ‘formulaic sequences’ (Wray, 1999; Erman and Warren, 2000; Wray, 2002; Schmitt, 2004) as prefabricated multi-word units. For the type-I pattern of TELL, formulaic sequences that can be frequently found in ICE-GB include, for example, (I) (’ll) tell you what (9), tell me this/that (3) and ...was telling me that... (3).49 Without anticipating the discussion of this issue in section 3.4, it should be noted that the notion of pattern frame as a systematisation of frequent lexical realisations of a pattern sits uneasily on the fuzzy boundary between syntactic creativity and formulaic routine in language use. For type Ie, the indirect object again turns out to be strongly biased towards personal pronouns in general (84 of 97 cases = 86.6%). The abundance of such referential items is symptomatic of the fact that the specification of the corresponding semantic role, i.e. the affected entity, is very often as unnecessary as in type I because its identification can be recovered from the context.50 There is also a clear tendency for language users to choose four kinds of wh-clauses in direct-object position, namely wh-clauses introduced by what (43), whether/if (14), how (13) and why (9). These four types account for 81.4% of all 97 instances of the type-Ie pattern. The conventionalised lexicalisation of the indirect and the direct object in the type-Ie pattern of TELL is summarised in (221). (221)
Ie (S) TELL [Oi:NP] [Od:wh-clause] frequent wh-clauses in Od-position (79 of 97 cases = 81.4%): what... (43), whether/if... (14), how... (13), why... (9) personal pronouns (84 of 97 cases = 86.6%): you (34), me (28), us (12), them (5) him (4), her (1)
48 Note, however, that about in the final slot, introducing a postmodifying prepositional phrase, usually co-occurs with specific nouns as NP heads in direct-object position (which are marked by an asterisk in Figure 3-5). 49 These and other formulaic sequences involving TELL, which often function as ‘conversational routines’ (Aijmer, 1996) are also listed in the Longman Grammar (cf. Biber et al., 1999: 753, 1002, 1008). Note, by the way, that while the concept of pattern frame is not bound to a specific word-form, formulaic sequences very often are: I tell you what usually cannot be changed to ?I told you what; nor is ?he tells you what acceptable. 50 The clear preference for personal pronouns in many TELL-patterns distinguishes TELL from GIVE. As will be pointed out later in this section, this may well have to do with the fact that TELL is a typically ‘spoken verb’.
Aspects of Description
139
In light of the lexical items that are frequently used as direct or indirect object in type Ie, it is possible to deduce a pattern frame from the findings summarised in (221). This pattern frame, which visualises the restricted lexical range of the typeIe pattern of TELL, is given in Figure 3-6.51
...
TELL
Figure 3-6:
you me us them him ...
what whether/if how why wh-
A pattern frame for the type-Ie pattern of TELL
In type Ig, pronominalisation of the indirect object is a significant factor again, but clearly less forceful than in the patterns described above. Personal pronouns account for 15 out of 26 instances (= 57.7%) of type Ig. Unlike type I, there is no discernible trend towards specific lexical items in the to-infinitival realisation of the direct object in type Ig. From a semantic point of view, the type-Ig pattern of TELL is used in negative contexts much more frequently than all other TELLpatterns, negativity being created either by obvious negation (e.g. ...not to attend, ...not to drink too much, ...not to report for work) or by using lexical verbs with a more or less negative connotation (e.g. ...to bury the community charge, ... to get out of town, ...to keep away from her). Disregarding some borderline cases, these negative and semi-negative instances account for half of all occurrences of the type-Ig pattern in ICE-GB (13 of 26 cases = 50.0%). This is not surprising, given the semantic role of the transferred entity in type Ig, which has been described as MESSAGE/ORDER. In comparison with the other TELL-patterns, it thus stands to reason to consider type Ig as displaying a neutral to negative ‘semantic prosody’ (Louw, 1993). The two factors that are relevant to the selection of the typeIg pattern of TELL are summarised in (222). Since pronominalisation of the indirect object is a relatively weak factor in this particular case and due to the lack of clearly identifiable lexical routines in the direct object, a pattern frame cannot be suggested on the basis of the ICE-GB data.52
51
Note that in Figure 3-6, all other realisations of the direct object in the type-Ie pattern of TELL are covered by ‘wh-‘ because, by definition, the wh-clause is introduced by a whword. 52 The conclusions given in (222) have to be taken with a measure of caution since they hinge on a relatively thin database. Notwithstanding this caveat, it seems as though type Ig is the only TELL-pattern which is more or less evenly distributed in spoken (15) and written language (11), while all other TELL-patterns are much more frequent in spoken than in written language. At the end of this section, it will be argued that not only the strikingly different overall frequencies of the TELL-patterns but also the difference in their distribution across spoken and written language can be functionally explained.
140 (222)
Chapter 3 Ig (S) TELL [Oi:NP] [Od:to-infinitive] personal pronouns (15 of 26 cases = 57.7%)
neutral (13) or negative (13) semantic prosody (50.0% vs. 50.0%)
The discussion of the four most frequent TELL-patterns should be concluded with a brief remark on the obvious relevance of heaviness and focus considerations. Since in all four patterns at hand (i.e. I, Ie, If, Ig) there is a tendency for the indirect object to be realised as a personal pronoun, it is quite clear that the subsequent constituent, i.e. the direct object, is very often heavier (especially in its clausal realisations in types Ie, If and Ig) than the indirect object. Also, the direct object almost always represents new information (to be placed in focus position), while the pronominalised indirect object is given and marked by a very low degree of ‘communicative dynamism’ (Firbas, 1992: 6ff.) from an information-structural point of view. The type-IPf pattern is the passive pattern that can be derived from its active counterpart If. To be precise, the indirect object becomes the subject so that the affected entity is now in initial position. The transferred entity, i.e. the verbal message realised as a that-clause, remains in situ, and the acting entity may be realised as an optional by-agent in final position. In a similar vein to the passive GIVE-patterns (see section 3.1.1.2), the by-agent is usually omitted (31 of 37 cases = 83.8%).53 The most important factor that leads language users to choose type IPf is the optionality of the by-agent, which makes it possible to leave out the acting entity. As shown in (223), the affected entity (i.e. the subjectivised indirect object) is again very often realised as a personal pronoun (25 of 37 cases = 67.6%), thus representing given information that has already been made explicit or that can be inferred from the situational context. What is more, in 7 other cases (18.9%) the passive subject is a definite noun phrase or a general noun and can also be regarded as given information since it has already been ‘textually evoked’ or ‘situationally evoked’ (Prince, 1981).54 Examples (224) and (225) show that the previous activation of the affected entity may lead to a 53
Note, however, that in all six cases in which the by-agent is explicitised it is in fact placed before the direct object, e.g. in The waiting cameramen were told by her son that she was sad not bitter and in But we’re told by John Redwood that all this is old hat . This movement of the by-agent may be motivated by the attempt to make it clear that the by-agent belongs to TELL and not to the verbal process of the that-clause. In the following sentence, for example, the by-agent may be regarded as either belonging to the verbal process of PRESENT (more probable) or to TELL (less probable): To her surprise and her sisters’ annoyance Pig was told in 1931 that she would be presented at Court by Lady Hammond-Graeme, who made her living out of the daughters of the nouveaux riches . 54 Even in most of the remaining cases, the indefinite noun phrase in subject position can be easily subsumed under ‘given information’ on the grounds of the situational context. For example, the indefinite plural item MPs is evoked by the preceding item Parliament in: ... when the 1989 Dock Work Act was presented as a Bill of Parliament, MPs were told that it would probably lead to 1,500 to 2,000 redundancies ... .
Aspects of Description
141
coordinated sentence structure in which TELL shares a syntactic slot with another verb: in (224), people... is the affected entity of both SIT UPON and TELL, while I in (225) is at the same time acting entity of PHONE and affected entity of TELL. Examples (226) and (227) illustrate that not only the givenness of the initial element but also the heaviness (and newness) of the final element, i.e. the that-clause in direct-object position, is relevant to the selection of type IPf. In fact, in 35 of 37 cases (94.6%) the direct object is significantly heavier than the passive subject.55 (223)
IP f [S < Oi active] BE told [Od:that-clause] (by-agent) given/activated before (32 of 37 cases = 86.5%): personal pronoun (25), definite NP/evoked (7)
heavy/new left out (35 of 37 cases (31 of 37 cases = 94.6%) = 83.8%)
(224) No new prophets are emerging. People who do come forward claiming to be prophets are promptly sat upon and told that this is something shameful this is audacious (225) I phoned Butterworths today and was told that, as we are a major client of yours (and a poor underfunded University Library) it is likely that you would take pity on us and supply us with these missing pages free of charge (226) Uh he would be told that the reason is that they have appraised the use of the resources and have redeployed them in a way which they consider to be more effective ... (227) The jury was told the event went off well and there was no trouble until it was winding up shortly before 3 am In the type-III pattern, the indirect object is omitted so that the affected entity of the TELLING event remains unspecified. As in the type-III pattern of GIVE (cf. (108) on p. 107), specific lexical items in direct-object position seem to trigger off the preference for this particular pattern. In 14 of 27 cases (51.9%) it is story/stories. Additionally, truth occurs three times (11.1%) and lies and facts occur twice (7.4% each) in the direct-object position of the type-III pattern of TELL. As a matter of fact, all these items can be subsumed into one semantic field, to which also many other items that are only attested once in this pattern in ICE-GB (e.g. jokes, tales) could easily be added. These findings suggest that, firstly, specific – and semantically similar – lexical items as realisations of the transferred entity function as lexical pointers to the type-III pattern and that, 55
The only exceptions in ICE-GB are provided by (224) and the following example: The report claimed that the LDA and other local enterprise companies set up by Scottish Enterprise had been told that only 68 percent of the total money they had asked for was available .
142
Chapter 3
secondly, the restricted set of lexical items in direct-object position is linked to leaving the affected entity unspecified because its mere existence (i.e. a recipient of the verbal message) is already implied in the lexical meaning of the ditransitive verb TELL. In fact, in all 27 cases, the affected entity can either be recovered from the context or its specification is irrelevant in a given context. What is more, the association between the type-III pattern of TELL and story/stories is so strong that the sequence of ‘TELL + story/stories’ should be best regarded as a lexically stabilised idiom. The same holds true for the sequence ‘TELL + difference + between’ because it is based on (and implies) a metaphorical extension of literal TELL.56 In (228) to (231), the lexical restrictions on the type-III pattern of TELL are summarised and exemplified. The examples are also intended to illustrate a significant tendency for this pattern to be used in negative contexts, namely in 16 of 27 cases (59.3%). Such negative contexts may be based on clear negation, items with a negative connotation (in the TELL-pattern itself or in its vicinity) and/or the description of unplesant processes or phenomena. (228)
III (S) TELL [Od:NP] Oi contextually recoverable / specification irrelevant (all 27 cases = 100.0%) recurrent lexical items (≥2): story/stories (14), truth (3), lies (2), facts (2), difference between... (2) (23 of 27 cases = 85.2%)
(229) But according to the United Bible Societies these figures don’t tell the whole story as some countries imported paper on which to print their own Bibles (230) And there are those of us who feel that perhaps this was going on a bit too obviously overboard in telling the story and setting the situation in a way which favoured the cricket (231) As the reader, we question whether Bishop Lowth is telling the truth, or is it all false and ‘imaginary’ as the author politely suggests? In the type-IV pattern, neither the direct object nor the indirect object is explicitised. Again, this pattern is primarily linked to specific idiomatic phrases and to less central meanings of TELL that could be seen as metaphorical extensions of literal TELL (in the sense of the meaning groups ‘have clear effect’ 56
Accordingly, the Macmillan English Dictionary (Rundell, 2002: 1476), for example, subsumes this phrase into a separate meaning group of TELL, which is described as ‘recognize sth’. In this meaning group, which also embraces the phrasal verb ‘TELL so./sth. FROM so./sth.’, no affected entity can be specified, while this is possible for ‘TELL + story/truth/lie/fact’ etc., which are subsumed under the meaning group ‘give information’.
Aspects of Description
143
and ‘fail to keep secret’ as specified in the Macmillan English Dictionary, cf. Rundell, 2002: 1476). The fact that we only find 11 instances of this pattern of TELL in ICE-GB seems to corroborate the strong semantic and pragmatic restrictions on type IV. In (232), all 11 instances are listed. (232)
IV (S) TELL Oi Od no specification of Oi and Od because 1. the pattern is linked to metaphorically extended meanings: (a) ‘have clear effect’ as in racing experience here will tell, too early to tell for certain; (b) ‘fail to keep secret’ as in don’t tell, how can they tell, you could tell, he was telling, we can never tell 2. the pattern is used in idiomatic phrases: as far as I can/could tell... (3), as Slater tells...
As shown in Figure 3-3 (p. 128), the type-V pattern, in which the transferred entity is not explicitised, is the second most frequent pattern of TELL in ICE-GB. It is hardly surprising that the direct object is omitted (by using the type-V pattern) whenever there is no need for any specification of the verbal message, either because it can be recovered from the context or because it is the process of telling as such that is important (and not the particularities of what is told). In this pattern, there is again a strong preference for pronominal forms both in the subject and in the indirect-object slot. In fact, personal/reflexive pronouns can be found in 89 of all 116 (grammatical and/or notional) subjects (76.7%) and in 93 of all 116 indirect objects (80.2%). Also, it is not unusual for the type-V pattern to follow the conjunction as (10 of 116 cases = 8.6%). These findings, which are summarised in (233), suggest that the type-V pattern is very often not used to introduce new bits of information to the discourse, but rather as a secondary comment, as an add-on (especially to direct speech) or a time-buying phrase (which is important for speakers under real-time constraints in spontaneous conversation in order to hold the floor). While contextual recoverability of the direct object is exemplified in (234) and (235), the comment, add-on and timebuying functions of the type-V pattern are illustrated in (236), (237) and (238) respectively. (233)
V (S) TELL [Oi:NP] Od contextually recoverable / specification irrelevant (all 116 cases = 100.0%) pronominalisation (93 of 116 cases = 80.2%): you (37), me/myself (29), us (8), them (7), him (6), her/herself (6) pronominalisation (89 of 116 cases = 76.7%): I/me (40), you (29), she (7), he (6), we/us (3), they (3), it (1)
144
Chapter 3
(234) And I had to tell them because they can’t suddenly say you’ve forgotten your room number or anything can you ... (235) I would much prefer to tell you one-to-one (236) And, as I told you before, should you need a hand with anything, be sure to call me up (237) ‘I’m afraid I’ve always been bad at names,’ she told him for she’d no recollection of him (238) Uh and then as you told us his uh h he he was interested in uh a business which was within a twenty-five mile radius of Hammersmith From (233), a pattern frame can be derived for the type-V pattern of TELL, which is sketched out in Figure 3-7.
... as
Figure 3-7:
I/me you she he ...
TELL
you me/myself us them him her/herself ...
A pattern frame for the type-V pattern of TELL
It should be noted in passing that the lexicogrammar of TELL is quite different from GIVE in that TELL is very often used with the indirect object alone, while the second most frequent pattern of GIVE is type-III in which only the direct object is explicitised (cf. Table 3-1, p. 99). One could easily hypothesise, then, that there is a stronger syntactic link between TELL and its indirect object than to its direct object, while for GIVE the reverse would hold true. In a sense, it seems as though from a quantitative point of view the indirect object is the ‘primary object’ of TELL (while the direct object is the ‘secondary object’ of TELL), whereas it is the ‘secondary object’of GIVE (while the direct object is the ‘primary object’ of GIVE).57 From a cognitive perspective, however, the corpus findings suggest that it is the semantic role of the direct object (i.e. the transferred entity) that is particularly strongly associated with TELL – and more inherent in the situation type evoked by TELL – because it remains unspecified at the syntactic surface structure much more frequently than the indirect object. Finally, the type-VP pattern should be briefly mentioned. It is the passive counterpart of type V, and it occurs only 11 times in ICE-GB. Examples have 57
Note that my use of ‘primary object’ and ‘secondary object’ is linked to specific ditransitive verbs in English, while Dryer (1986) applies these terms in order to compare the syntax of different languages. In fact, he introduces the notions of primary and secondary objects in order to distinguish between different kinds of direct and indirect objects from a typological perspective.
Aspects of Description
145
already been given in (197) to (199). As shown in (239), this pattern is marked by pronominalisation in the passive subject, and the by-agent is omitted in all 11 cases. In fact, this pattern makes it possible to explicitise only the affected entity of the TELLING event since the transferred entity can be recovered from the context and a specification of the acting entity is considered unnecessary. (239)
VP [S < Oi active] BE told Od (by-agent) pronouns (10 of 11 cases = 90.1%)
left out (all 11 cases = 100.0%) contextually recoverable (all 11 cases = 100.0%)
Although all of the nine most frequent TELL-patterns are marked by preferred lexical realisations of specific clause elements, there are three general variables that should be taken into account in order to describe why language users choose one particular pattern in a given context: (1) the need to explicitise the acting entity, the affected entity and/or the transferred entity; (2) the kind of transferred entity (if explicitised); (3) the need to nominalise the verbal message inherent in the transferred entity (if explicitised) in order to present the underlying FACT or ANSWER as a PRODUCT. In fact, the influence of the three variables on the process of pattern selection can be systematised along the lines of a flowchart in which the three variables come into play at particular stages; see Figure 3-8. These three variables alone allow for a unified explanation of the routinised pattern selection for TELL. These three context-independent variables therefore seem to be much more central to the pattern selection than the context-dependent factors that were described for each individual pattern of TELL. While each individual pattern may display its specific routines in actual language use, all frequent TELL-patterns can be systematised at a more abstract and decontextualised level according to the three semantic variables listed above. In this context, note that the flowchart given in Figure 3-8 includes no information on the context-dependence of specific semantic roles (as could be derived from, say, the extent to which a clause element tends to be pronominalised in a pattern) nor on specific frequent lexical items that may be regarded as lexical pointers to a pattern. Also, it should not go unmentioned that heaviness and newness of clause elements are not taken into account in Figure 3-8. Generally speaking, from the analysis of the TELL-patterns above the conclusion has to be drawn that heaviness and newness may be communicatively important factors but they are clearly less distinctive principles of pattern selection for TELL than for GIVE. As mentioned before, it is, for example, telling that in more than 60% of all occurrences of TELL (i.e. in the patterns I, Ie, If and Ig) there is no difference whatsoever in terms of the arrangement of the semantic roles, while no such clear-cut preference for one particular arrangement can be found for GIVE. The variables that are mentioned in Figure 3-8 may help to explain why specific patterns are more frequent than others in terms of the number and kind of semantic roles involved: note that particular options in Figure 3-8, namely the ones marked by an asterisk,
146
Chapter 3
Transferred entity to be explicitised? YES *
NO
*
Affected entity to be explicitised?
Affected entity to be explicitised?
YES *
YES *
Acting entity to be explicitised?
NO
type III
Acting entity to be explicitised? YES *
NO
YES *
NO
NO
type IV
type VP
type IPf type V
What kind of transferred entity? * FACT type If
*
* ANSWER type Ie
*
ORDER type Ig
To be presented as PRODUCT? Figure 3-8:
*
type I
Routinised pattern selection: the case of TELL
indicate preferred pathways when it comes to the verbalisation of a TELLING event: the type-I, type-Ie, type-If and the type-V pattern. This frequency-based line of argumentation will be taken up in chapter 4, which deals with the cognitive implications of corpus-based findings on the usage of ditransitive verbs. I would like to conclude the discussion of TELL by briefly referring to the preferred use of TELL in spoken language. In particular, while GIVE is more or less evenly distributed in the spoken and written section of ICE-GB (see section 3.1.1.3), more than two thirds of all occurrences of TELL (438 of 636 instances = 68.9%) can be found in the spoken component. Given the fact that the spoken component in ICE-GB is significantly larger than the written component, it is appropriate to normalise the number of occurrences in terms of frequencies per million words. Table 3-9 gives normalised frequencies of each routinised TELLpattern in the spoken and written section of ICE-GB. The figures reveal that TELL, unlike GIVE, is a typically ‘spoken verb’. This is also corroborated by the Longman Grammar (cf. Biber et al., 1999: 367f.), in which GIVE is described as evenly distributed across conversation, fiction writing, news writing and academic prose, while TELL is clearly biased towards spoken language (including the written representation of spoken language, say, in fiction writing). The prevalence of TELL in spoken language may be one important reason why in many
Aspects of Description Table 3-9: TELLpattern type I type Ie type If type Ig type IPf type III type IV type V type VP sum
147
Frequency of TELL-patterns in the spoken and written section of ICE-GB spoken section
written section
number of occurrences
frequency (pmw)
number of occurrences
frequency (pmw)
total
76 67 140 15 26 19 9 78 8
119.2 105.1 219.5 23.5 40.8 29.8 14.1 122.3 12.5
20 30 75 11 11 8 2 38 3
47.2 70.8 177.0 26.0 26.0 18.9 4.7 89.7 7.1
96 97 215 26 37 27 11 116 11 636
TELL-patterns context-dependent pronominal realisations of clause elements prevail (cf. Biber et al., 1999: 1042ff.). Also, it is against this background of TELL as a typically spoken verb that the surprising distribution of the type-Ig pattern can be explained. Note that this pattern is the only one which occurs more frequently in written than in spoken language (cf. Table 3-9). This pattern, however, is marked by a to-infinitive in direct-object position which represents a specific kind of verbal message, namely a MESSAGE/ORDER (cf. Figure 3-4, p. 134). It may be not too far-fetched an assumption that “the general atmosphere of cooperativeness” (Stenström, 1994: 1) in spoken conversation puts some restrictions on the explicit statement of ORDERS by using the type-Ig pattern: TELL tends to be used to refer to (or report on) a directive speech act, but is rarely used performatively. In a sense, then, the avoidance of overt ORDERS in spoken language could be regarded as yet another, more pragmatic principle of pattern selection for TELL. 3.2
Habitual ditransitive verbs in ICE-GB
As sketched out in Figure 2-4 (p. 84), habitual ditransitive verbs are defined in the present study as those verbs that occur fairly frequently in ICE-GB (i.e. 200 to 700 occurrences), but that are not strongly associated with an explicit ditransitive syntax (i.e. not parsed as ditransitive in ICE-GB) in the clear majority of all cases (cf. Table 2-5, p. 84). These two defining aspects render them less typical of ditransitivity than GIVE and TELL from a lexicogrammatical point of view, although their basic meaning is in line with the semantically three-valent core of ditransitivity (cf. Figure 1-8, p. 36). In ICE-GB, four verbs turn out to be habitual ditransitive verbs: SHOW, ASK, SEND and OFFER. Table 3-10 lists the total number of occurrences of these verbs in ICE-GB (including phrasal/prepositional
148
Chapter 3
verbs) and the relative percentage of explicit ditransitive syntax for each of the four verbs. The figures in the last column show that it is only in a minority of cases – ranging from some 13% to some 27% – that each of the four verbs is used in an explicit ditransitive syntax. Table 3-10: Habitual ditransitive verbs in ICE-GB verb SHOW ASK SEND OFFER
total number of occurrences 639 518 346 198
explicit ditransitive syntax occurrences percentage 84 13.1 % 91 17.6 % 79 22.8 % 54 27.3 %
The following sections provide an overview of important principles of pattern selection for SHOW (see section 3.2.1), ASK (see section 3.2.2), SEND (see section 3.2.3) and OFFER (see section 3.2.4). 3.2.1 SHOW SHOW is associated with a variety of patterns, most of which have already been discussed for GIVE (see section 3.1.1.1) and TELL (see section 3.1.2.1). However, some SHOW-patterns are attested neither with GIVE nor with TELL. In this context, there are also some analytical problems of demarcation that need to be addressed. Both the new patterns and the problems of categorisation will be discussed first in this section. Not infrequently, SHOW is followed by a noun phrase in direct-object position and an ing-clause, as in (240) and (241): (240) And so this little diagram shows uh light rays coming down from the sun (241) Uhm the red line on the other hand shows heat going back out away from the earth into space … With these sentences, multiple analyses are possible. Specifically, the question arises as to whether these sentences should be categorised along the lines of a complex-transitive analysis (i.e. as SVOC) or of a ditransitive analysis (i.e. along the lines of the type-III pattern). Note, for example, that the Longman Grammar (cf. Biber et al., 1999: 391) subsumes sentences such as But they still show boardroom salaries growing faster than middle management’s under complex transitive complementation rather than ditransitive complementation because the ing-clause is seen as a complement to the direct object (i.e. boardroom salaries). As a matter of fact, this analysis is absolutely plausible; but it is also possible, in my view, to analyse this very sentence differently so that it could still be regarded as a variant of ditransitive complementation. The two analyses of this sentence
Aspects of Description
149
are given in (242). Analysis I is the one suggested by Biber et al. (1999) in which the ing-clause is regarded as a separate clause element. Analysis II is the one preferred in the present study because the ing-clause is seen as a postmodification of the preceding noun phrase (analogous to, say, a relative clause) within the clause element of direct object. In analysis I, SHOW is thus taken to be complextransitive (‘cxtr V’), while analysis II is based on the assumption that SHOW continues to be a ditransitive verb (‘ditr V’), although the affected entity is not made explicit in the form of an indirect object. (242) I:
S
cxtr V
direct object
complement to direct object
But they still show boardroom salaries growing faster than middle management’s II:
S
ditr V (indirect object) direct object type-III pattern of SHOW
With SHOW, unlike TELL, it is possible to have a wh-clausal and that-clausal realisation of the direct object in a type II-pattern. These patterns are coded ‘IIe’ and ‘IIf’ respectively. In these patterns, the indirect object is realised as a tophrase which is followed by a clausal direct object. Formulas and examples are given in (243) to (246). (243) IIe (S) SHOW [Oi:PPto] [Od:wh-clause] (244) The prophet responds to this by saying that God will show to them all too clearly how just he is by coming against them in judgement (245) IIf (S) SHOW [Oi:PPto] [Od:that-clause] (246) And you have to consider whether the plaintiff has shown to you that the words do mean what he or she uh suggests Another SHOW-pattern that is not attested with GIVE or TELL, is type IIIPrg. An example of this pattern is given in (247). (247) In theories about “sexual politics”, the personal sexual relationship is shown to hold overtones of domination and submission, with generally the male dominating the female As the pattern code ‘IIIPrg’ indicates, the syntax of the SHOW-pattern as exemplified in (247) is structurally related to a passive form derived from the active type-III pattern. What is special about this pattern, however, is the fact that the passive form is based on a ‘raising’ (‘rg’) of the object from a subordinate to-
150
Chapter 3
infinitive clause to the superordinate clause (cf. Quirk et al., 1985: 1202f.). The analytical steps involved in the ‘raised object’ analysis of (247) may be taken to start off from the following hypothetical version with a type-I pattern: (247a) …they show [us]Oi [that the personal sexual relationship holds overtones of…]Od (=> type I) In a second step, it may be assumed that the affected entity is no longer explicitised as the indirect object, and that the that-clause is changed into a to-infinitive clause with its own notional subject (i.e. the personal sexual relationship), which is at the same time the syntactic object of SHOW: “semantically, it has the role of subject of the nonfinite verb; but syntactically it is ‘raised’ from the nonfinite clause to function as object of the superordinate verb” (Quirk et al., 1985: 1202). This is the case in (247b), and in analogy to the preferred analysis II in (242) the underlying SHOW-pattern is categorised as type III. (247b) …they show Oi [the personal sexual relationship]raised obj [to hold overtones of…]to-inf. (=> type III) In a third step, the raised object becomes the subject of the passive, resulting in the passive pattern IIIPrg, cf. (247) above. On the face of it, this pattern is no longer related to the transfer event of the basic version in (247a). From a semantic point of view, however, the type-IIIPrg pattern may still be taken to display a ditransitive meaning and represent an abstract transfer event: the initial element may well be regarded as the affected entity, realised as the passive subject, which ‘is given the description or the attributes’ as specified in the to-infinitive clause (which may thus be re-interpreted as the transferred entity in this context). In other words, the entity in the to-infinitive clause is figuratively transferred to the entity in the passive subject. In (247) above, for example, the personal sexual relationship may thus be taken to represent the affected entity which ‘is given a specific description’ (i.e. to hold overtones of…). This interpretation can be mapped onto the pattern formula of the type-IIIPrg pattern as follows: (248)
IIIPrg [S < Od raised] BE shown [to-infinitive] (by-agent) ‘is given the following description:’
The semantic reinterpretation of the type-IIIPrg pattern as outlined above makes it possible to link this SHOW-pattern to the ditransitive meaning in terms of an underlying figurative transfer event. Among the type-IV patterns, we also find a peculiar SHOW-pattern that warrants a brief discussion of its inherent semantics. Apart from the usual type-IV pattern, in which only the acting entity is explicitised (while the affected entity and the transferred are omitted), SHOW also occurs in a pattern which looks
Aspects of Description
151
structurally similar but is semantically different. This additional SHOW-pattern is coded ‘IVerg’ because SHOW can be considered an ergative verb in this pattern: “With ergative verbs, the relationship between the object and the verb in the SVOd [i.e. type-III] pattern is the same as that between the subject and the verb in the SV [i.e. type-IVerg] pattern” (Biber et al., 1999: 147). Semantically, the grammatical subject in type IVerg corresponds to the direct object in the type-III pattern. From a semantic point of view, then, it is the transferred entity which is realised as the grammatical subject in type IVerg, while in type IV the grammatical subject is also the notional subject, referring to the acting entity.58 This semantic difference between the type-IVerg and the type-IV pattern of SHOW is shown in (249) to (252).59 Note that because the ergative use of a verb such as SHOW precludes any grammatical object, there is no indication of object deletion (i.e. ‘Oi’ and ‘Od’) in the pattern formula of the type-IVerg pattern in (249). (249) IVerg [S ergative] SHOW (250) Uh I’m afraid this does not show very well graphically (251) IV (S) SHOW Oi Od (252) The response of Joyce and Proust to that severed connection, as Beckett shows, lies in the totalising dynamic which both deploy as a means of giving their writing some foundation or necessity Finally, brief mention should also be made of the type-IVP pattern, which is the passive form with no objects and an optional by-agent (which, however, is always omitted in ICE-GB), see (253) and (254).60 58 It should not go unmentioned that Biber et al.’s (1999) use of the term ‘ergativity’ is not very common, although I find it suitable for the purpose at hand. Usually, ergativity is a typological label ascribed to specific languages to distinguish between nominative-accusative and ergative-absolutive systems, which differ in case-marking and in the unmarked choice of argument roles for particular clause elements (cf. Dixon, 1979; Du Bois, 1987; Dixon, 1994; van Valin, 2001). Nevertheless, I keep to Biber et al.’s (1999) languageinternal understanding of ergativity because it captures the fundamental difference in the argument structure of type IV and type IVerg (although they share the same surface structure). Note that some other terms are also used for the ergative use of verbs, e.g. ‘middle construction’ (Fagan, 1988) and ‘mediopassive’ (Hundt, 2001). 59 In a sense, the use of SHOW in type IVerg resembles the specific meaning of TELL when used as the adjective telling: and I think actually it’s quite telling because if you were to do that and add it up ... . Here, the semantic relation between the subject it and telling as an adjective is similar to the one between the direct object and TELL as a verb. 60 As will be discussed later in this section, one may think of two different active counterparts, i.e. either with the omitted by-agent as a subject or with the diagram as an inanimate subject (the diagram shows…).
152
Chapter 3
(253) IVP shown Oi Od (by-agent) (254) When returning the wide serve, try to follow the path of the arrows, as shown here in the diagram As with GIVE and TELL, all phrasal and prepositional verbs including SHOW, e.g. SHOW AS, SHOW THROUGH and SHOW UP, have been excluded from the corpus analysis for semantic reasons. Thus, 606 instances of SHOW proper are left. Table 3-11 provides an overview of the total number of occurrences of all SHOW-patterns in ICE-GB and their relative frequencies. Table 3-11: Frequency of SHOW-patterns in ICE-GB type I Ib Ie If
pattern
(S) SHOW [Oi:NP] [Od:NP] [Od: NPantecedent] (rel. pron.) [S] SHOW [Oi:NP] (S) SHOW [Oi:NP] [Od:wh-clause] (S) SHOW [Oi:NP] [Od:that-clause] miscellaneous IP [S < Oi active] BE shown [Od:NP] (by-agent) IPe [S < Oi active] BE shown [Od:wh-clause] (by-agent) II (S) SHOW [Od:NP] [Oi:PPto] IIb [Od:NPantecedent] (rel. pron.) [S] SHOW [Oi:PPto] IIe (S) SHOW [Oi:PPto] [Od:wh-clause] IIf (S) SHOW [Oi:PPto] [Od:that-clause] IIP [S < Od active] BE shown [Oi:PPto] (by-agent) IIPb [antecedent]co (S< Od)co (BE) shown [Oi:NPto] (by-agent) III (S) SHOW [Od:NP] Oi IIIa (S) SHOW [Od:wh-clause] Oi IIIb [Od:NPantecedent] (rel. pron.) [S] SHOW Oi IIIc (S) SHOW [Od:that-clause] Oi miscellaneous IIIP [S < Od active] BE shown Oi (by-agent) IIIPb [antecedent]co (S < Od)co (BE) shown Oi (by-agent) IIIPrg [S < Od raised] BE shown [to-infinitive] (by-agent) miscellaneous IV (S) SHOW Oi Od IVerg [Sergative] SHOW IVP shown Oi Od (by-agent) V (S) SHOW [Oi:NP] Od sum
sum 43 9 14 6 9 2 1 12 1 1 1 0 4 218 33 13 110 8 31 39 14 2 7 10 10 8 606
% 7.1 1.5 2.3 1.0 1.5 0.3 0.2 2.0 0.2 0.2 0.2 0.0 0.7 36.0 5.4 2.1 18.2 1.3 5.1 6.4 2.3 0.3 1.2 1.7 1.7 1.3 100
The overall quantitative trends in Table 3-11 are in line with the frequencies that can be found for the various valency patterns of SHOW in the Longman Gram-
Aspects of Description
153
mar (cf. Biber et al., 1999: 390). For example, this holds true for the high frequency of type III (corresponding to ‘SVOd(NP)’), the fairly frequent use of types IIIa and IIIc (corresponding to ‘SV+complement clause’) and the low frequencies of type II (i.e. ‘SVOdOi(PrepP)’) and types Ie and If (i.e. ‘SVOi+ complement clause’).61 In Figure 3-9, the twelve most frequent patterns are given. These twelve patterns, which will be discussed in detail below, account for more than 90% of all instances of SHOW. The most frequent pattern, accounting for more than one third of all occurrences of SHOW, is the type-III pattern. Thus, it is perfectly common to omit the indirect object of SHOW, leaving the affected entity of the underlying event type unspecified.
36.0%
18.2%
7.1% 6.4%
9.7% 5.4% 5.1%
Ty pe Ty III pe III c Ty Ty pe I pe III P Ty b pe III Ty a pe III Ty P Ty pe I pe e III P Ty rg pe III b Ty p Ty eI pe I IV Ty erg pe IV P O th er s
2.3% 2.3% 2.1% 2.0% 1.7% 1.7%
Figure 3-9:
The twelve most frequent SHOW-patterns
The type-III pattern is the default pattern of SHOW on quantitative grounds. Not only is it characterised by the omission of the indirect object, but also by the realisation of the direct object (i.e. the OBJECT being shown) as a noun phrase, as shown in (255). The examples given in (256) to (257) illustrate that this default pattern is available both for animate and inanimate acting entities (see underlined subjects). 61
As mentioned before, an in-depth comparison between the ICE-GB data as analysed here and the Longman Grammar is difficult for two reasons: there is no one-to-one correspondence between the valency patterns in the Longman Grammar and the patterns distinguished in the present study, and the Longman Grammar does not indicate exact frequencies, but only distinguishes between six frequency ranges (see section 3.1.2.1).
154
Chapter 3
(255) III (S) SHOW [Od:NP] Oi (256) I’m going to show some illustrations today from two quite different earthquakes ... (257) Figures for the region as a whole show a marked discrepancy between expectation of life at birth in the larger wealthier states and the smaller poorer ones Table 3-12 shows that there is a discernible preference for the type-III pattern in written genres. Table 3-12: Frequency of the type-III pattern of SHOW in the spoken and written section of ICE-GB type-III spoken section pattern number of frequency (per of instances mill. words) SHOW 105 164.7
written section number of frequency (per instances mill. words) 113 266.7
sum total 218
It turns out that language-users’ choice of other patterns than the type-III pattern is very often related to one or more of the above-mentioned three features that are typical of the default pattern of SHOW: (1) no specification of the affected entity; (2) realisation of the direct object, representing the transferred entity, as a noun phrase; (3) a bias towards written language. Starting off from the type-III pattern as the default pattern of SHOW, one of the basic questions that arises concerning the pattern-selectional process is not – unlike typical ditransitive verbs – why the affected entity is omitted, but rather why it is explicitised in specific contexts. The second most frequent SHOW-pattern is type IIIc. Here, the direct object is realised as a that-clause. Capitalising on what has been said about the meaning of that-clauses in section 3.1.2.2, this pattern is chosen whenever the transferred entity is an OBJECT/FACT (rather than an OBJECT/PRODUCT as in type III). What is more, this pattern tends to be used whenever the acting entity is inanimate. Specifically, the lexical items in the subject position of the type-III pattern can be subsumed under the semantic field of scientific DATA (e.g. opinion poll, survey, result). Related to this semantic field is the use of items that refer to scientists themselves (e.g. Stückli (1980), scientist) or to historical events as data (e.g. Gulf crisis, this century). As shown in (258), this semantic restriction on the subject position of the type-III pattern is a major quantitative factor (80 of 110 cases = 72.7%). Some examples are given in (259) to (261). On the other hand, personal pronouns can be found in only 19 of 110 cases (= 17.3%), but even these instances very often refer to the extended semantic field of scientific DATA because the personal pronoun in these contexts refers to a scientist (or a group of scientists) producing – or reporting on – such data, see (262).62 62
The pronoun it in subject position is excluded here although the context makes it clear that it usually refers to a previously mentioned noun that belongs to the semantic field of
Aspects of Description (258)
155
IIIc (S) SHOW [Od:that-clause] Oi OBJECT/FACT referring to the extended semantic field of DATA (80 of 110 cases = 72.7%): 1. scientific and other data: e.g. opinion poll(s) (7), survey (5), result(s) (4), study/-ies (4), evidence (3), facts (3), figures (3), research (3) 2. scientists: e.g. researchers (1), scientist (1), Stückli (1980) (1), supervisor (1) 3. historical events/developments: e.g. Gulf crisis (1), history (1), this century (1)
(259) The Harris poll shows Mr Heseltine is the most popular with Mr Major second (260) A scientist named Bridgeman claimed to have shown that there was a deviation from Ohm’s law at very high densities ... (261) This century has already shown that there are times when right demands might, and this will be one of them (262) We ’ve shown at the National Institute that if you take an industry which uh can be found in both uh say France uh Germany and Britain that the presence of a skilled workforce in Germany has a very obvious uhm effect on the products which that country can produce In principle, two similar principles of pattern selection are involved in the choice of the type-IIIa pattern in which the direct object is realised as a wh-clause. The wh-clause is used to present a transferred entity which could be described as OBJECT/ANSWER (in a similar vein to the semantic role of MESSAGE/ ANSWER with which the type-Ie pattern of TELL is associated, see section 3.1.2.2). In 29 of 33 cases (= 87.8%), the wh-clause is introduced by how or what. Also, there is a clear preference for inanimate subjects in this pattern, and many of the lexical items occurring in the subject position of type IIIa can again be subsumed into the extended semantic field of DATA. The two semantic restraints on the type-IIIa pattern are shown and exemplified in (263) to (265). (263)
IIIa (S) SHOW [Od:wh-clause] Oi OBJECT/ANSWER: how... (20), what... (9) referring to the extended semantic field of DATA (22 of 33 cases = 66.7%): e.g. diagram (2), example (2), story/-ies (2), survey (1), water test (1)
DATA, e.g. map: If you look at a map, it shows you that you don’t know the area ... .
156
Chapter 3
(264) This example clearly shows how the choosing and presentation of artefacts can totally not only misrepresent people, but also carries messages of power and opportunities for exploitation (265) although this diagram aims to show what varying boundaries can do, it shows effectively the way slight variation can totally alter the appearance of a map In type IIIb, the direct object is realised as a noun phrase (as in type III), but is verbalised as the first element and functions as an antecedent to which a relative clause, including SHOW and its subject, refers back. As in the case of, say, the type-Ib pattern of GIVE (cf. (89) in section 3.1.1.2), the semantic role of transferred entity can be shown to be fronted because it is part of a preceding pattern and, thus, serves as the first element of the SHOW-pattern, see (266). The choice of the type-IIIb pattern of SHOW is another example of the influence that the pattern flow in natural language exerts on language users’ selection of a specific pattern in a given context. As shown in (267) to (268), for example, the type-IIIb pattern of SHOW is chosen because its initial element (e.g. the form, the interest) is already part of the preceding pattern in the text (e.g. wish so. back to..., thank you for...). Note also that in 9 of 13 cases (= 69.2%) the subject of the relative clause is a personal pronoun. (266)
IIIb [Od:NPantecedent] (rel. pron.) [S] SHOW Oi part of a previous pattern (all 13 cases = 100.0%)
personal pronoun (9 of 13 cases = 69.2%)
(267) They are wishing him back to the form that he showed throughout the nineteen-eighties (268) I would like to thank you again for the interest you have shown in our company The passive pattern IIIP, in which the direct object of type III becomes the subject, is similar to the active type IIIb in that the transferred entity is in initial position while the acting entity is placed towards the end – preceding SHOW in type IIIb and following SHOW in type IIIP. As has been noted above, the transferred entity is in initial position in type IIIb because it is part of a preceding pattern. In type IIIP, on the other hand, the transferred entity is in initial position because it provides thematic information that has already been activated (i.e. situationally or textually evoked) before, see (269). What renders this passive pattern of SHOW particularly interesting is the fact that the acting entity is usually not omitted. In 5 of 31 cases (=16.1%) a by-agent is used. What is more, in another 16 cases (= 51.6%) the semantic role of acting entity, i.e. the entity that is showing something, is not realised as a prepositional phrase introduced by by but by in (14) or on (2). This is possible because the acting entity in the type-IIIP pattern is usually inanimate, e.g. figure, character, graph, map. As the illustrative
Aspects of Description
157
examples in (270) to (272) reveal, it is possible to change them into sentences with the corresponding active type-III pattern, regardless of whether the acting entity is realised as a by-phrase or as a prepositional phrase introduced by in or on. Semantically, then, this pattern is an exception to the usual ‘valence-decreasing’ quality of the passive (Dryer, 1986: 819). Since the acting entity is made explicit in type IIIP in the clear majority of all cases, it is reasonable to assume that it is usually not the optionality of the by-agent that makes language users choose this pattern, but rather the opportunity of placing the acting entity in final focus position. In essence, two complementing and information-structural principles of pattern selection can be identified for the type-IIIP pattern: (1) the initial element (i.e. the transferred entity) presents given information and is therefore introduced as the first element, and (2) the acting entity refers to new information and is thus placed in end-position, realised as a prepositional phrase introduced by by, in or on.63 (269)
IIIP [S < Od active] BE shown Oi (by-agent) activated before/ taken up (all 31 cases = 100.0%)
ACTING ENTITY: made explicit as prepositional phrase introduced by by, in or on (21 of 31 cases = 67.7%)
(270) However, Carolingian forms of government were indeed being extended into central and Western Brittany, this is shown by the appearance of vicecomites => ... the appearance of vicecomites shows this (271) It uses the traditional Mascot buffered Channel in a test network. The Mascot 3 network is shown in Fig. 1 => ... Fig. 1 shows the Mascot 3 network (272) The principle was once again to superimpose a radio frequency current on a much larger direct current and to measure the the the radio frequency resistance of the gold leaf as the direct current was varied. The result of this very careful series of measurements is shown on the next on this graph => ... The next/this graph shows the result of this very careful series of measurements In type IIIPb, the transferred entity is not only shifted to front position, but it also serves as an antecedent for a subsequent relative clause or a past participle construction, see (273). Like the structurally similar active type-IIIb pattern (see above), this pattern is chosen whenever other patterns in the text require a shift of 63 In (270) to (272), dotted underlining is intended to mark the information to which the transferred entity of the SHOW-pattern refers back. The SHOW-pattern itself is underlined, and the active counterpart is given in italics: this alternative version shows that the prepositional phrase represents the semantic role of acting entity.
158
Chapter 3
the transferred entity to preverbal front position. In most cases the transferred entity again turns out to be at the same time part of a preceding pattern (26 of 39 cases = 66.7%). This is illustrated in (274) and (275), where the video and the address are part of the preceding patterns remember sth. and send sth. to somewhere. But pattern flow may also come into play somewhat differently in the selection of the type-IIIPb pattern. In 7 of 39 cases (= 17.9%), the first element of the anticipated pattern to the right, e.g. sth. reveals that... as in (276) and sth. possesses sth. as in (277), is also presented as the transferred entity of SHOW so that the type-IIIPb pattern is used in order to have the transferred entity function as an antecedent to which a subordinate relative clause or past participle construction can be added. (273)
IIIPb [antecedent]co (S < Od)co (BE) shown Oi (by-agent) part of a previous pattern (26 of 39 cases = 66.7%) and/or part of a subsequent pattern (7 of 39 cases = 17.9%)
(274) I remember for instance the video that was shown of the lorry going across the bridge (275) About a fortnight before your vehicle licence expires, DVLA will send a reminder form B11, or V85/1 in the case of some heavy goods vehicles, to the address shown on the Registration Document (276) A survey of 150 feature films made in the US between 1938 and 1974, and shown on British TV revealed that the cry ‘Let’s get outta here!’ was used once in 84% of them and more than once in 17% (277) The Von Neumann computer – shown in simple block form in Fig 1a – possesses a central possessor attached to memory By introducing a new clausal structure, the type-IIIPb pattern helps to reduce the decoder’s processing load in a specific way. From a more cognitive point of view, it has been argued that each clause tends to have one specific thematic ‘focus’ (cf. e.g. Chafe, 1979). It could well be argued that the type-IIIPb pattern of SHOW is used because it is easier for the recipient to process two separate clauses with two individual foci. The fact that in many cases – as in examples (274) to (277) – SHOW is complemented with a prepositional phrase (providing, for example, information on the inanimate acting entity of the SHOWING event) seems to indicate a relatively high cognitive load. By using SHOW in its type-IIIPb pattern, this cognitive load is to some extent reduced. Note, for example, that by omitting shown in (275), Registration Document would not be presented as a new ‘focus’ but as part of the already complex and semantically loaded main clause. If, on the other hand, SHOW is used, the recipient is able to finish off his/her processing of the first ‘focus’ (i.e. DVLA will send ... to the address) before
Aspects of Description
159
having to start processing the second ‘focus’ (i.e. shown on the Registration Document). The syntax and semantics of the type-IIIPrg pattern have already been discussed before, see (248) above. To recapitulate, I would contend that from a strictly semantic point of view sentences such as (279) and (280) also display the inherent ditransitive meaning of SHOW. In other words, the basic meaning of ditransitive verbs as visualised in Figure 1-8 (p. 36) can also be mapped onto the type-IIIPrg pattern of SHOW: the passive subject is the affected entity which is given a description as specified in the to-infinitive (i.e. the transferred entity). As shown in (278), the type-IIIPrg pattern of SHOW is chosen whenever the affected entity represents, in the broadest sense, given information, that is to say the affected entity is placed in initial position because it is anchored in the preceding text, i.e. situationally and/or textually evoked by a previously mentioned entity or process. Accordingly, it can be argued that students are evoked by universities in (279), and women and men by feminist in (280). Note that usually the acting entity seems to be irrelevant: in 13 of 14 cases (= 92.9%) there is no by-agent. (278)
IIIP rg [S < Od raised] BE shown [to-infinitive] (by-agent) given/evoked (all 14 cases = 100.0%)
‘is given the following description:’
left out (13 of 14 cases = 92.9%)
(279) In the other universities abroard abroad the students are shown to have more life within them (280) If Nellie prefigures the late twentieth-century feminist, then women and men are shown to have equal potential for perverse wielding of power The type-I pattern is similar to the default type-III pattern of SHOW in that the direct object is also realised as a noun phrase. Type I is different from type III with regard to the explicitation of the affected entity as the indirect object. At first sight, one might hypothesise that the type-I pattern is used if the affected entity is not recoverable from the context and/or when its specification is important. However, it should be noted that in 39 of 43 cases (= 90.7%), the indirect object is a personal pronoun, see (281). As the examples in (282) to (283) illustrate, it would be easily possible to omit the indirect object altogether because the affected entity would be implicitly understood anyway. (281)
I (S) SHOW [Oi:NP] [Od:NP] personal pronoun (33 of 43 cases = 74.4%): I (22), he (4), they/them (2), she (2), we (1), you (1), it (1)
personal pronoun (39 of 43 cases = 90.7%): you (25), me (4), us (4), him (3), her (3)
160
Chapter 3
(282) I ’ll just quickly show you one or two more (283) And what I want to do is I want to kind of show you the way in which we construct some of these uhm phonematic uhm and exponency issues within the sythesizer at York In this context, it is important to note that the type-I pattern turns out to be a ‘spoken pattern’: 36 of 43 instances (= 83.7%) can be found in the spoken section of ICE-GB. As has been noted for various TELL-patterns before (see section 3.1.2.2), language users seem to explicitise the indirect object in spoken language not because it is by any means semantically salient and/or contextually necessary, but rather in order to establish a cooperative atmosphere. This is especially true of the frequent use of you in the indirect-object position of type I. Although in all contexts at hand the personal pronoun could be omitted without jeopardising the transaction of the message, it is usually not omitted in order to keep in line with the medium-dependent cooperative demands and the need to constantly refer to the shared context (cf. Biber et al., 1999: 1042). Personal pronouns also abound in the subject position of the type-I pattern. It is clear from the data given in (281) that very frequently the type-I pattern is lexically realised as ‘I SHOW you ...’. In virtually all these cases, however, an additional (modal or non-modal) verb (e.g. will, am going to, intend to) and/or an adverb (e.g. just, quickly) is inserted between I and SHOW. It seems as though the sequence ‘I SHOW you...’ is considered too direct and, perhaps, too confrontational in most spoken contexts, so that additional material before the verb may function as a softener and help reduce the confrontational quality of the meaning of ‘I SHOW you ...’. This is exemplified in (282). In (283), want to and the discourse marker kind of might fulfil a similar face-saving function, but probably the speaker is still planning the sentence here and is buying himself some time. The type-I pattern is very useful in this context because it makes it possible for him to introduce the ‘general noun’ way (cf. Halliday and Hasan, 1976) which serves as a peg on which to hang, as it were, the description of what he actually wants to show. Otherwise, he might have used the type-Ie pattern: I want to show you how we construct... would be an alternative phrasing without a general noun. The fact that the type-I pattern is much more frequent in spoken than in written language may thus have to do with its potential to allow for an extended speech-planning process under on-line constraints. The aforementioned type-Ie pattern occurs only 14 times with SHOW. Apart from the frequent pronominalisation of the indirect object (12 of 14 cases = 85.7%), which could be systematised along the lines of the pattern frame suggested for TELL in Figure 3-6 (p. 139), no such discernible trend is observable for the subject position. Also, there are no clear routines in the realisation of the direct object. As has been pointed out in section 3.1.2.2, the wh-clause in the directobject position of type Ie fulfils the semantic role of MESSAGE/ ANSWER. It is thus not surprising that the type-Ie pattern is much more frequently attested with TELL as a verb of verbal communication than with SHOW which, in its literal meaning, refers to the transfer of (the impression of) a physical object. In fact, the
Aspects of Description
161
type-Ie pattern of SHOW is linked to a figurative meaning of SHOW, described by the Macmillan English Dictionary as ‘give instructions etc.’ (Rundell, 2002: 1319). In other words, in the type-Ie pattern SHOW is used as a verb of verbal communication and can thus easily be replaced by, say, TELL. As shown in (284), the meaning of the wh-clause impinges on the meaning of the ditransitive verb and imparts a specific meaning to the pattern in its entirety. In (285) to (286), two examples of the type-Ie pattern of SHOW are given. (284)
Ie (S) SHOW [Oi:NP] [Od:wh-clause] MESSAGE/ANSWER ‘give instructions etc.’ Ie (S) SHOW ≈ TELL [Oi:NP] [Od:wh-clause]
(285) It’s not an easy thing to describe in words but any good rose-growing manual will show you how to do it => ... any good rose-growing manual will tell you how to do it (286) For example uh does forty-nine show us what was happening there => ... does forty-nine tell us what was happening there The type-II pattern, in which the indirect object is realised as a to-phrase in final position, is relatively rare with SHOW: only 12 occurrences can be found in ICEGB, see (287). As in the case of the type-II pattern of GIVE (cf. (98) on p. 104), one important factor that is responsible for language users’ choice of the type-II pattern of SHOW is the use of the pronoun it in direct-object position. One example is given in (288). With it referring to the transferred entity, the type-I pattern is usually not used (e.g. (?)I’ll show you it later). This is, to a lesser extent, also true of this in direct-object position, see (289). In the remaining cases, focus and/or weight considerations seem to be primarily responsible for the selection of the type-II pattern of SHOW. In other words, the direct object is placed in initial position because it refers to a previously activated entity and represents given information (which would, of course, pertain to it and this as well). The indirect object is placed in end-position because it represents new information which is often specified by a postmodifying relative clause (rendering the constituent heavier than the direct object) and/or which the language user is going to elaborate on in the subsequent text. An illustrative example, taken from a conversation (between speakers A and B), is given in (290).
162 (287)
Chapter 3 II (S) SHOW [Od:NP] [Oi:PPto] new and/or heavy (all 12 cases = 100.0%) given information: it (3), this (2) or other items referring to previously activated entities (all 12 cases = 100.0%)
(288) I mean I ’ll show it to you later (289) If the Department of Social Security has sent you a certificate AG 2 you should show this to the optician at the time of your sight test (290) B: So at the moment uh disabled dancers are still sh shut out of the dance world Uhm they may be allowed into dance therapy but why should they want to study dance therapy more than an able-bodied dancer A: Right Let’s move on to uhm what independent dance or dance of this nature offers the other activities which would normally be associated with the d disabled people what more what do you think this offers B: I think I’d like to answer that in a s in a slightly different way Uhm I showed uh a sp a a video of some of the work we’d done to some of my college students who are on a dance course And uhm their response was uhm very positive uhm Note that in (290) speaker B, in speaking of a video of some of the work, refers back to a previously activated topic, namely his work on the development of a dance therapy for disabled persons (in boldface). The transferred entity thus clearly represents given information. On the other hand, some of my college students is in focus position because this is the information that is new to the discourse and on which speaker B intends to elaborate further. Accordingly, he specifies the particular students that he is referring to by the relative clause who are on a dance course and takes up this newly introduced entity in the subsequent clause by describing their response (in italics). The type-IVerg pattern is also rarely attested. Obviously this has to do with the fact that in this pattern SHOW displays a particular and untypical meaning, which the Macmillan English Dictionary describes as ‘be able to be seen’ (Rundell, 2002: 1319). For convenience, the syntactic structure of this pattern is repeated in (291). As has been pointed out at the beginning of this section, the grammatical subject here does not represent the acting entity but rather the transferred entity. For example, it is easily possible to rephrase the examples given in (292) to (293) in the form of the default type-III pattern by making explicit the acting entity as the grammatical subject and realising the transferred entity as the direct object. The type-IVerg pattern, thus, turns out to be used whenever language users see no compelling reason for making explicit the acting entity or the affected entity. Since the transferred entity is shifted to the unusual subject position, the verb as such and the pattern in its entirety are imparted with a specific meaning. In this particular regard, the type-IVerg pattern is similar to the
Aspects of Description
163
type-Ie pattern of SHOW (cf. (284) on p. 161) in that a semantic principle of pattern selection is primarily responsible for its use. (291)
IVerg [Sergative] SHOW TRANSFERRED ‘be able to be seen’ ENTITY IVerg [Sergative] SHOW ≈ can be seen/noticed
(292) Uh I’m afraid this does not show very well graphically => I’m afraid the graph does not show this very well (293) Now make sure the green neutral light is showing => Now make sure it is/you are showing the green neutral light Finally, the type-IVP pattern should be mentioned. From a structural point of view, it is the passive form that is related to the active type-IV pattern. However, the active-passive correspondence suggested here is somewhat different from the correspondences that the Comprehensive Grammar describes for so-called ‘-ed participle clauses’ (cf. Quirk et al., 1985: 994f.). Notwithstanding this difference, I subscribe to Quirk et al.’s (1985: 994) general analysis of this construction as “both syntactically and semantically passive”. As shown in (294), the type-IVP pattern of SHOW is to a large extent routinised since in 9 of 10 cases (= 90.0%) it occurs in the form of as (has been) shown. As has already been noted for the type-IIIP pattern, the semantic role of acting entity is often realised as an adverbial, e.g. as in Contents list in example (295) and here in example (296). Note, however, that while it is possible to change (295) to an active sentence by means of a simple rearrangement of items, this is not possible for (296) since here is not admissible in subject position.64 A genuine by-agent is not attested with the typeIVP pattern in ICE-GB. Although the type-IVP pattern is strongly routinised in the form of as (has been) shown, it still allows for variation: in (297), we find where shown in lieu of as shown. Thus, it is important to take into account the type-IVP pattern as a syntactic pattern of SHOW, although it almost always occurs in the formulaic sequence as (has been) shown. (294)
IVP shown Oi Od (by-agent) as (9 of 10 cases = 90.0%)
64
ACTING ENTITY: left out (5 of 10 cases = 50.0%) or made explicit as adverbial (5 of 10 cases = 50.0%)
Of course, one could easily think of other active versions of the sentences at hand that are based on the proxy pronoun it functioning as the subject: …as it shows in the ‘Contents’ list (295), …as it shows here (296), where it shows (297).
164
Chapter 3
(295) The rules are explained in more detail under their own section headings (as shown in ‘Contents’ list) => ... as the ‘Contents’ list shows (296) This analysis can be displayed graphically as shown here => *... as here shows (297) Crease only where shown The discussion of SHOW-patterns in this section has again revealed that different principles of pattern selection are responsible for language users’ preference of one pattern over others in a given context and that individual patterns are marked by different kinds and degrees of routine. In principle, all the factors that were identified for the routinised pattern selection of GIVE and TELL (cf. Figure 3-2, p. 111; Figure 3-8, p. 146) have been shown to be relevant to the selection of SHOW-patterns as well: • • • • •
influence of preceding patterns and the anticipation of subsequent patterns; lexical items as pointers to a specific pattern; focus and weight considerations; explicitation and arrangement of semantic roles; clause elements as representations of specific subtypes of semantic roles.
It needs to be emphasised that a clear-cut factor loading for each individual factor is difficult to calculate for three reasons: (1) many factors often come into operation along with each other (e.g. end-focus and end-weight) and it is difficult to prioritise one factor over another in such cases; (2) sometimes one specific factor may imply another factor (e.g. the pronoun it as a lexical pointer implying that the subsequent constituent is almost always heavier); (3) most important, the decision on which factors seem to play a significant role in a specific context is, in the final analysis, a matter of linguistic interpretation of the data, which is very often based on world knowledge and experience that cannot be found in the text itself. Despite the overall plausibility of the lexical, semantic, pragmatic and textual principles of pattern selection suggested for SHOW in this section, it turns out to be virtually impossible to base them on a solid statistical basis, say, by carrying out a linear discriminant analysis (cf. e.g. Gries, 2001). While such statistical tests are very useful for the comparison of a limited number of alternative constructions, factors and linguistic data, it is very difficult to apply them successfully to a wide range of lexicogrammatical patterns, a multitude of possibly significant principles of pattern selection and the entirety of a 1-million-word corpus. For example, Gries (2003) is certainly right in making a plea for a ‘multifactorial analysis of syntactic variation’ such as the choice between John picked up the book and John picked the book up or between John gave him the book and John gave the book to him. However, it is difficult to analyse a large corpus exhaustively on a multifactorial basis because the relevant instances of the two variants have to be coded for the various factors by hand. Also, the binary choice that is in
Aspects of Description
165
focus in many studies of the dative alternation only covers a restricted range of the patterns in which a given verb (e.g. GIVE) is used. I would contend that a focus on the alternation between the type-I and the type-II pattern of, say, GIVE does not provide a comprehensive picture of the patterning of this verb in actual usage. Finally, note that the general quantitative trends that have so far been presented in this chapter are considered plausible not primarily because of their statistical significance but because it is possible to explain them in functional terms; in Mair’s (forthcoming) terminology the principles of pattern selection that have been suggested so far should therefore be seen as ‘linguistically significant’. It goes without saying that this approach combines corpus-based methodology with an intuition-based analysis of corpus data (see section 2.1). In the following section, this traditional kind of corpus-based approach will be applied to the ditransitive verb ASK. 3.2.2 ASK The verb ASK is attested 410 times in ICE-GB. The frequency of all ASKpatterns in the corpus is given in Table 3-13 below. Again, phrasal and prepositional verbs such as ASK ABOUT, ASK FOR and ASK someone OUT have been excluded from the corpus analysis because their semantics differ from ASK. The focus here is on the lexicogrammar of the ditransitive verb ASK. The eleven most frequent patterns are given in Figure 3-10; they include 368 instances of ASK and make up for 89.8% of all occurrences of the verb in ICE-GB. It is these eleven patterns that the following discussion of principles of pattern selection for ASK will focus on. There is one pattern in Table 3-13 that has not yet been discussed before, which is the type-IIId pattern. In this pattern, as in all other type-III patterns, the indirect object is omitted. The direct object is realised as a to-infinitive: (298) They may they may ask to see that just out of interest Note also that in Table 3-13 no type-II pattern is listed for ASK, i.e. no pattern in which the indirect object is realised as a prepositional phrase in final position. This is in stark contrast to the Comprehensive Grammar which draws an analogy between the usual to-phrase for GIVE, TELL, SHOW and other ditransitive verbs on the one hand and an of-phrase for ASK on the other:65
65 The same correspondence is also suggested by Klotz (2000: 178) for favour. However, he explicitly points out that with permission or asylum (e.g. in ?She first had to ask permission of her boss and ?The opposition leader asked asylum of the U.S.) the of-construction is dubious. Huddleston and Pullum (2002: 311) state that “[w]ith ask the of version is somewhat more formal, and unlikely where the direct object expresses a ‘concealed question’ ... : He asked me my name/the time ∼ ?He asked my name/the time of me.”
166
Chapter 3
Table 3-13: Frequency of ASK-patterns in ICE-GB type I Ib Ie Ig IP IPe IPg III IIIa IIIb IIIc IIId IIIP IIIPb IV V VP sum
pattern (S) ASK [Oi:NP] [Od:NP] [Od: NPantecedent] (rel. pron.) [S] ASK [Oi:NP] (S) ASK [Oi:NP] [Od:wh-clause] (S) ASK [Oi:NP] [Od:to-infinitive] miscellaneous [S < Oi active] BE asked [Od:NP] (by-agent) [S < Oi active] BE asked [Od:wh-clause] (by-agent) [S < Oi active] BE asked [Od:to-infinitive] (by-agent) miscellaneous (S) ASK [Od:NP] Oi (S) ASK [Od:wh-clause] Oi [Od:NPantecedent] (rel. pron.) [S] ASK Oi (S) ASK [Od:that-clause] Oi (S) ASK [Od:to-infinitive] Oi miscellaneous [S < Od active] BE asked Oi (by-agent) [antecedent]co (S < Od)co (BE) asked Oi (by-agent) miscellaneous (S) ASK Oi Od miscellaneous (S) ASK [Oi:NP] Od miscellaneous [S < Oi active] BE asked Od (by-agent)
sum 16 8 50 90 6 1 3 33 2 18 45 5 1 6 7 2 1 11 34 3 63 2 3 410
% 3.9 2.0 12.2 22.0 1.5 0.2 0.7 8.0 0.5 4.4 11.0 1.2 0.2 1.5 1.7 0.5 0.2 2.7 8.3 0.7 15.4 0.5 0.7 100
Occasionally, a preposition other than to and for occurs in this function: She asked Paul a favour. ∼ She asked a favour of Paul. (Quirk et al., 1985: 1211) However, this particular example of favour – and the inherent correspondence between the type-I and the type-II pattern suggested here – hides a systematic semantic difference between the two sentences. While favour refers to what she wants in she asked Paul a favour, it represents what Paul does in she asked a favour of Paul. This difference becomes obvious if favour is replaced with, say, question. It is possible to use question in the type-I pattern but it is not admissible to change it into a seemingly corresponding type-II pattern, see (299). With advice as direct object, on the other hand, the of-phrase is possible, but not the typeI pattern, see (300).
Aspects of Description
167
22.0%
15.4% 12.2% 11.0%
10.1%
8.3% 8.0% 4.4% 3.9%
Ty pe I
g Ty pe V Ty pe Ty I e pe III Ty a pe Ty IV pe IP g Ty pe III Ty pe Ty I pe Ty I b pe II Ty I d pe III b O th er s
2.0% 1.5% 1.2%
Figure 3-10:
The eleven most frequent ASK-patterns
(299) She asks him a question => *She asks a question of him (300) She asks advice of him => *She asks him advice The syntactic correspondence between type I and type II that is suggested in the Comprehensive Grammar on semantic grounds is therefore questionable. As the examples in (299) and (300) show, it is much more plausible from a semantic point of view to treat ASK OF as a prepositional verb in its own right and to refrain from establishing a type-II pattern for ASK by drawing an analogy between the of-phrase of ASK and, say, the to-phrase of GIVE. The most frequent pattern of ASK in ICE-GB is the type-Ig pattern, in which the direct object is realised as a to-infinitive, see (301). It may be structurally more complex than the basic type-I pattern but should nevertheless be considered the default pattern of ASK not only for quantitative but also for underlying semantic reasons. The meanings of ASK that are associated with ASK in the type-Ig pattern are described in the Macmillan English Dictionary as ‘expect sth’, ‘say you want sth done’ and ‘invite sb to do sth’ (Rundell, 2002: 68). In examples (302) to (304), examples are given for these three basic meanings. On the other hand, the meanings that are associated with the type-I pattern are given as ‘try to get information’ and ‘tell sb you want sth’.
168
Chapter 3
(301) Ig (S) ASK [Oi:NP] [Od:to-infinitive] (302) I ’d ask the tenant to either rebuild or underpin the wall (≈ expect sth.) (303) I was about to go when Keith asked me to stay (≈ say you want sth. done) (304) and I think they were just sort of making up for for the embarrassment of having to ask me such an experienced artiste to come in for those two lines (≈ invite sb. to do sth.) It is reasonable to assume that the meanings of ASK associated with the type-Ig pattern are more central to the verb because they are much more frequently attested. In other words, it is more common for ASK to be combined with a direct object that refers to the semantic role of MESSAGE/ORDER (realised as a toinfinitive and covering the full range from polite requests to commands) than with a direct object that refers to the semantic role of MESSAGE/QUESTION (realised as a noun phrase or, alternatively, as a wh-clause in type Ie). Not only the fact that different subtypes of semantic roles are associated with different patterns of ASK is something that is reminiscent of what has been found for TELL (see section 3.1.2.2).66 Note also that all type-I patterns together account for more than half of all occurrences of ASK in ICE-GB. As with TELL (cf. Table 3-7, p. 127), it is thus usually the case that both the transferred entity and the affected entity are realised as objects of ASK. There are also other quantitative trends that TELL and ASK share, for example the frequent use of the typeV pattern (16.4% for TELL and 15.4 % for ASK), the similar frequency of the type-III pattern (3.8% for TELL and 4.4% for ASK) and the prevalence of one particular passive pattern (5.2% for the type-IPf pattern of TELL and 8.0% for the type-IPg pattern of ASK). Both TELL and ASK are verbs of verbal communication and instantiations of the situation subtype of SPEAKING within the ditransitive semantics (cf. Figure 1-8, p. 36). It seems not too far-fetched an assumption that this semantic similarity may be partially responsible for the similar quantitative trends in the lexicogrammar of the two verbs. Starting off from the default type-Ig pattern, the selection of the type-Ie pattern, which is the third most frequent ASK-pattern, is based on the need for language users to present the transferred entity not as a MESSAGE/ORDER but as a MESSAGE/QUESTION, for which a wh-clause provides an option.67 In 66 Note that in the type-Ig pattern – as in all other patterns of ASK – the indirect object of ASK represents a semantic role that is similar to the affected entity of TELL: for both verbs the indirect object typically refers to the recipient of a verbal message (while the indirect object of GIVE and SHOW usually refers to the recipient of a physical or visual object). 67 The lexical meaning of ASK implies that the transferred entity is interpreted as QUESTION, while the lexical meaning of TELL imparts the same semantic role with the quality of ANSWER (see section 3.1.2.1).
Aspects of Description
169
effect, the question that is being put forward is verbalised as a wh-clause in direct-object position. The affected entity continues to be realised as an indirect object. As shown in (305), there is a strong tendency both for the subject and the indirect object to be pronominalised (78.0% and 72.0% of all cases respectively), which indicates their ‘given’ (i.e. thematic) status in most contexts and the language user’s reference to the shared context. That the shared context is utilised by way of pronominalisation to such a large extent has to do with the fact that the type-Ie pattern is a ‘spoken pattern’: of 50 instances, 42 can be found in the spoken section and only 8 in the written section of ICE-GB. As for the direct object, the same wh-clauses as in the type-Ie pattern of TELL (cf. Figure 3-6, p. 139) hold the field: if/whether-clauses representing yes/no questions (23), whatclauses referring to inquiries about factual descriptions (11), how-clauses representing questions about the mode of an action (6) and why-clauses asking for reasons (4). Some examples of these routines in the three slots of the type-Ie pattern are given in (306) to (308). Example (308) is also intended to show that usually the sequence I ask you wh-... seems to be considered too confrontational and is therefore hedged, as it were, with a modal verb or some other ‘filling’ material in between the subject I and ASK.68 There is one instance in ICE-GB in which nothing is inserted between the subject I and ASK. This sentence is given in (309); it strikes even the casual reader as rather impolite and face-threatening. Indeed, this is not surprising since it is taken from a legal cross-examination. (305)
Ie (S) ASK [Oi:NP] [Od:wh-clause] pronouns (39 of 50 cases = 78.0 %): I (21), you (7) he/she (6), we (3), they (2)
MESSAGE/QUESTION: whether/if... (23), what... (11), how... (6), why... (4) (44 of 50 cases = 88.0%) pronouns (36 of 50 cases = 72.0%): him/her (13), me (12), you (5), them (3), ourselves (3)
(306) I don’t think he ’s ever asked me what I was doing (307) I asked him how he interpreted Mr Hurd’s remark about economic and monetary union (308) I was going to ask you if you were ever going to send my tape back (309) Now I ask you if you could explain why it’s not until twelfth of January that you signed that application The quantitative trends as summarised in (305) are so strong that a pattern frame for the type-Ie pattern of ASK can be suggested, which is given in Figure 3-7. It 68
In this, the sequence I ask you wh-... is similarly constrained as the sequence I show you sth., see (278).
170
Chapter 3
includes the pronominalisation and wh-clausal routines involved in the use of this pattern. The asterisks refer to the fact that we in subject position and the reflexive pronoun ourselves tend to co-occur. I you he/she we* they ...
ASK
him/her me you them *ourselves ...
whether/if what how why wh-
Figure 3-11: A pattern frame for the type-Ie pattern of ASK In the type-I pattern, the direct object also represents the semantic role of MESSAGE/QUESTION, but the question is not made explicit in the form of a wh-clause. Specifically, while the type-Ie pattern presents the question as a process in action, the type-I pattern presents the question at hand as a product by way of nominalisation. Accordingly, Huddleston and Pullum (2002: 976) speak of ‘concealed questions’. Thus, a semantic distinction similar to the one made for the different type-I patterns of TELL (see section 3.1.2.2) should be drawn between the type-I patterns of ASK: • type-Ig pattern: • type-Ie pattern: • type-I pattern:
[Od:to-infinitive] represents the semantic role of MESSAGE/ORDER [Od:wh-clause] represents the semantic role of MESSAGE/QUESTION [Od:NP] represents the semantic role of QUESTION/PRODUCT
Both the clause elements [Od:NP] in the type-I pattern and [Od:wh-clause] in the type-Ie pattern refer to the semantic role of QUESTION. Thus, the semantics of the type-I pattern is strongly related to the type-Ie pattern, which is corroborated by the fact that the two patterns are associated with the same two meanings of ASK described above as ‘try to get information’ and ‘tell sb. you want sth.’. The two patterns are also lexically similar in that in type I, too, the subject and the indirect object tend to be pronominalised, see (310). But also in all the other cases in which pronouns are not used in the subject and/or indirect-object position, the lexical items are inferrable either from world knowledge or the context: people, the Secretary of State, and the speaker also refer to given information, see (311) and (312). What is introduced as new to the discourse is in fact the transferred entity, i.e. the question, in end-position. This question is not explicitised as a whclause but condensed, as it were, into a noun phrase, and the following three lexical items used as NP heads account for 14 of all 16 instances (87.5%) of the type-I pattern of ASK: question(s) (8), this (4), things (2). These lexical items may be used for two reasons. Either a question that has already been formulated
Aspects of Description
171
before is taken up or summarised in the direct-object position of the type-I pattern, or the fact that a question that is going to be formulated in the subsequent text is anticipated by the direct object of the type-I pattern.69 In both cases, the question at hand is thus condensed into a noun phrase in the type-I pattern because the pattern is not used to introduce the question itself but to speak about a question that is (to be) explicitised elsewhere. In (313), for example, this refers back to the question you don’t chop the text up, and in (314) the question anticipates the actual question which is still to follow (Whose view of...). (310)
I (S) ASK [Oi:NP] [Od:NP] given information (all 16 cases = 100%): I/me (8), other pronouns (5), other items (3)
QUESTION/PRODUCT => preferred items: question(s) (8), this (4), things (2) (14 of 16 cases = 87.5%) => explicitised question is taken up or anticipated given information (all 16 cases = 100%): you (8), other pronouns (5), other items (3)
(311) People always ask me this about my novels too (312) Can I can I ask the Secretary of State simply this (313) Now what you do is you build indexes / Is that right / you don’t chop the text up / I know I ’ve asked you this before but I’ve forgotten the answer (314) But before Conservative MPs follow any fevered rush to the brink of the political cliff, they should pause and ask themselves the question: Whose view of the Government’s Predicament and of its European policy, deserves their trust and support? In the type-Ib pattern, too, pronominalisation of the acting entity and the affected entity as well as the presentation of the transferred entity as a product come into play. In this pattern, however, the noun phrase in direct-object position is fronted and serves as an antecedent for a subsequent relative clause with ASK as the main verb. While the type-Ib pattern of GIVE has been shown to be chosen whenever the fronted element is part of a preceding pattern (see (89) on p. 102), this textual factor is of less relevance to the selection of the type-Ib pattern of ASK. Only in one example, which is given in (316), is there a preceding pattern (answer sth. to so./sth.) which provides the initial slot for the type-Ib pattern of ASK. In all other cases, however, as shown in (315), it turns out that the type-Ib pattern of ASK is primarily selected whenever language users want to highlight the issue of the question as such without formulating the question itself. This functional explan69
It is surprising that in Klotz’s (2000:177) analysis of this ASK-pattern, which is based on the COBUILD Corpus, question is not at all attested as a lexical item in direct-object position.
172
Chapter 3
ation, by the way, also accounts for example (316). In this context, it is interesting that seven of eight instances of this pattern are attested in the spoken section of ICE-GB (the only written example, which is given in (318), can be found in the genre of social letters). The conclusion can be drawn that the type-Ib pattern of ASK is a – mainly spoken – device by means of which language users shift their focus to the importance of a particular question which has already been mentioned before or which is going to be formulated in the subsequent text (or which is not to be specified at all). In a sense, fronting the transferred entity (i.e. the QUESTION/PRODUCT) in order to place special emphasis on this very entity runs counter to the principle of end-focus and is more in line with the contradictory principle of ‘task-urgency’ as suggested by Givón (1983): what is most important/urgent is verbalised first.70 The examples in (316) to (318) illustrate the highlighting effect of fronting the transferred entity by choosing the type-Ib pattern of ASK. Note that in examples (317) and (318) – unlike (316) – it would be easily possible to change the type-Ib pattern to the basic type-I pattern without having to change the preceding text. However, it is obvious that the frontposition of anything and so much is much more appropriate in these contexts. (315)
Ib [Od: NPantecedent] (rel. pron.) [S] ASK [Oi:NP] QUESTION/PRODUCT pronouns pronouns in front-position in order to (7 of 8 cases (all 8 cases be highlighted as important = 87.5%) = 100%) (all 8 cases = 100%)
(316) ... because she’ll answer yes to every question you ask her (317) Is it I mean once you’ve got the st more stuff about it if there’s anything you wanna ask me come back => ... if you wanna ask me anything come back (318) He may not even have liked his brother, and there is so much that I needed to ask him => ... and I needed to ask him so much There is one passive pattern among the eleven most frequent ASK-patterns, namely IPg. Here the indirect object of the corresponding active form (representing the affected entity) becomes the passive subject, and the direct object (representing the transferred entity and realised as a to-infinitive) remains in situ. As 70
Mathesius (1975) brings together the two principles and suggests that they account for different aspects of language use. He calls the unmarked order of elements (according to the principle of end-focus) the ‘objective order of theme before rheme’, while the rhemebefore-theme order (according to the principle of task-urgency) is defined as the ‘subjective order’ which is used in ‘emotionally coloured utterances’ (Mathesius, 1975: 156). Thus, both principles should be taken to be relevant mechanisms that underpin the order of elements in English.
Aspects of Description
173
in many other passive patterns of ditransitive verbs that have been discussed so far, the most important factor that leads language users to choose the type-IPg pattern of ASK is the optionality of the by-agent. In fact, in 29 of 33 cases (= 87.9%), the affected entity is not made explicit in the form of the by-agent, see (319) below. It should be mentioned in passing that in two of the remaining four cases, the by-agent is placed in the usual final position (e.g. in ...he would automatically be asked to be uh Prime Minister by the Queen), while in the other two cases the by-agent precedes the to-infinitive (e.g. in She was asked by the junior flying corps to come to the celebrations they were holding that night). This may suggest that once the by-agent is explicitised, the arrangement of the postverbal clause elements is not routinised but may be based on context-specific focus considerations. As for the majority of cases in which the by-agent is omitted, the question arises as to why it is the indirect object that becomes the passive subject (and is thus moved to front-position) and why it is the direct object that remains in end-position. Two complementary factors seem to play a role. First, the toinfinitive is not acceptable as a passive subject due to structural ‘object constraints’ (Quirk et al., 1985: 163f.). The Comprehensive Grammar gives an example of the monotransitive verb hope, but the same state of affairs pertains to ditransitive verbs as well: With clauses as objects, however, the passive transformation is to a greater or lesser degree restricted in use: ... John hoped to meet her. ∼ *To meet her was hoped (by John). (Quirk et al., 1985: 163) The structural inadmissibility of to-infinitival passive subjects in English may be hypothesised to be functionally motivated. That is, an underlying second reason for the structure of the type-IPg pattern, with the to-infinitive in the usual final position, may be the heaviness of the to-infinitival constituent in comparison with the passive subject. In 30 of 33 cases (= 90.1%), the to-infinitival direct-object clause is, not surprisingly, considerably heavier than the passive subject. Two illustrative examples are given in (320) and (321). (319)
IPg [S < Oi active] BE asked [Od:to-infinitive] (by-agent) / / [Od:to-infinitive] cannot function as passive subject
left out heavier than [Oi] (30 of 33 cases (29 of 33 cases = 90.1% = 87.9%)
(320) If the project is successful and a large number of ethnic minority families volunteer to adopt and foster, they may be asked to take in white children (321) You will be asked to put on the form the name and address of the local authority with whom you are registered
174
Chapter 3
(322) Why, then, was it not possible to appoint a team of three people from these candidates? Each of us would have happily accepted the Panel’s decision, but to appoint no-one from such highly qualified teachers was an insult to us all. How ironic that those of us who are C.S.Ts have been asked to produce materials for you demonstrating the best in modern practice in the field of specific learning difficulties! Examples (320) and (321), in which the passive subject is pronominalised, should, however, not lead to the conclusion that the passive subject tends to refer to given information and the to-infinitive to new information (according to the principle of end-focus). Rather, in the majority of cases, it is not easy – if not to say impossible – to decide on whether the information presented in the toinfinitive is ‘newer’ than that presented in the passive subject. As examples (320) and (321) show, it may well be that the to-infinitival direct object may refer to the only new bit of information in specific contexts, but the passive subject contains equally new information in as many other contexts. In (322) above, for example, those of us who are C.S.Ts can be regarded as new information to the same extent as the to-infinitive in the context at hand. To recapitulate, then, the most significant principle of pattern selection that is responsible for language users’ choice of the type-IPg pattern of ASK is the possibility to leave the acting entity unspecified. Additionally, the to-infinitive, which is usually the heaviest constituent, does not become the passive subject in the preferred passive pattern because toinfinitival passive subjects are not grammatically well-formed in English. It was shown in Table 3-13 (p. 166) that among the eleven most frequent patterns that account for 90% of all instances of ASK, four patterns can be found in which the indirect object is left out: III, IIIa, IIIb and IIId. Interestingly enough, these four patterns correspond to the four frequent active type-I patterns (i.e. I, Ib, Ie, Ig) that have been described above. The only difference between each of these type-I patterns and its type-III counterpart lies in the realisation or non-realisation of the affected entity as an indirect object. Table 3-14 provides a systematic overview of the corresponding type-I/type-III patterns and their frequencies. As has been noted for type-III patterns in general (i.e. in the discussion of GIVE, TELL and SHOW), the indirect object can usually be shown to be omitted either because the underlying semantic role, i.e. the affected entity, can be recovered from the context or because the specification of a particular affected entity is irrelevant in the context at hand. In a first approximation, it thus stands to reason that whenever these two factors come into play, language users may replace a type-I pattern with its type-III counterpart. The frequencies in Table 3-14, which are taken over from Table 3-13 (p. 166), reveal that the extent to which the affected entity is considered contextually recoverable – or its specification irrelevant – varies from pattern to pattern. Specifically, it turns out that the affected entity very often remains unspecified whenever the direct object is fronted (type Ib vs. type IIIb). On the other hand, if there is an explicitised question in the form of a wh-clause, language users seem to find it neccessary to specify the affected entity as the addressee of the question as well (type Ie vs. type IIIa). The same holds true for the
Aspects of Description
175
Table 3-14: Type-I and corresponding type-III patterns of ASK: formulas and frequencies type-I pattern formula and frequency type I (S) ASK [Oi:NP] [Od:NP] (3.9%) type Ib [Od: NPantecedent] (rel. pron.) [S] ASK [Oi:NP] (2.0%) type Ie (S) ASK [Oi:NP] [Od:wh-clause] (12.2%) type Ig (S) ASK [Oi:NP] [Od:to-infinitive] (22.0%)
type-III pattern formula and frequency type III (S) ASK [Od:NP] Oi (4.4%) type IIIb [Od:NPantecedent] (rel. pron.) [S] ASK Oi (11.0%) type IIIa (S) ASK [Od:wh-clause] Oi (1.2%) type IIId (S) ASK [Od:to-infinitive] Oi (1.5%)
[Oi] left out whenever the affected entity is recoverable from the context or whenever its specification is irrelevant in a given context use of ASK in combination with a to-infinitive which represents, broadly, the semantic role of MESSAGE/ORDER: again language users tend to specify the addressee of the order and do not therefore omit the indirect object too often (type Ig vs. type IIId). Notwithstanding these different quantitative trends, corpus data show that in all the instances of the type-III patterns listed in Table 3-14, the direct object is omitted because it is in fact unnecessary to specify the underlying semantic role in the contexts at hand. In effect, I suggest that each of the type-III patterns listed in Table 3-14 may be considered a ‘secondary form’ of its corresponding type-I pattern.71 The overriding principle of pattern selection that leads language users to prefer the type-III pattern to its corresponding type-I pattern is the contextual recoverability of the affected entity or the irrelevance of its specification. Examples (323) to (331) are intended to show that it is, in principle, possible to add an indirect object to each instance of any of the type-III, type-IIIa and type-IIIb patterns in order to identify and/or specify the affected entity (added in square brackets) of the verbal process.
71
As a matter of fact, this label can also be applied to type-I/type-III correspondences with other ditransitive verbs, e.g. GIVE, TELL and SHOW (see sections 3.1.1.2, 3.1.2.2 and 3.2.1). Note, by the way, that the label of ‘secondary form’ is used here in a non-technical sense and is not based on the concept of surface-structural movements as in generative grammar. Nor does it rule out that any type-III pattern may be routinised in its own particular ways, see for example the lexical routines involved in the type-III pattern of GIVE as summarised in (108).
176
Chapter 3
[type I] => type III (323) And unfortunately there isn’t a me… representative from the Arts Council with us tonight so we can’t ask [him/her] the question direct (324) What I wonder however is whether we could take it further and ask [ourselves] the more difficult question about this thing called secularisation (325) My theory was that the councils would read it and write me less often for advice / In fact they simply asked [me] harder questions